Modification history

This document describes incremental modifications made to the Baker lab software suite. Since the image preprocessing, image reconstruction, format conversion, and utility programs are so tightly coupled (e.g. shared standards and libraries) and released as a single package, all modifications are described together in a single document. Owing to the complexity of the software, it is impractical to provide a line-by-line listing of alterations. Instead we focus primarily on modifications that impact the end user and on large-scale changes to the code. Starting on version 4.02, more details on the changes in the latest version can be found in the file MINOR_RELEASE_NOTES, included in the software package.

• prj: calculate projections from a reconstruction using orientations either random, uniform or from an input data set
• a3dm_icosrotate: rotate an icosahedral map from standard orientation to one with a pair of vertices positioned along the z axis
• a3dm_dat_stats: gather information on the results from a run at a given iteration
• a3dm_sel_stableparts: select only stable particles, i.e. with unchanged orientation with respect to a previous iteration, and generate a new stack parameter file
• a3dm_plot_score: generate a histogram of the particle scores at a given iteration
• a3dm_mg_inspect_prepare: generate auxiliary info and images for screening micrographs using a3dm_mg_inspect
• a3dm_mg_inspect: quickly inspect and screen micrographs by means of a graphical interface
• avemaps: compute average of input maps (replaces mergemaps and mergemaps2)

• improved handling of cyclic and dihedral symmetries
• added program a3dm_snapshot to gather information about a run, to send to developers for troubleshooting
• particles read from stack files are now background-subtracted

• use inner radius given as input to setup_rmc to mask reference maps at each iteration
• suppressed non significant warnings

• added option po2r mask_map to specify a binary mask to use for mask-focused refinement
• added option auto evod_parallel to specify if processing of even- and odd-numbered particles to be done concurrently
• added option p3dr cpu_sparse to specify if to spread cpus among different nodes to optimize resources for large maps (works only on multi-node clusters with pbs/sge queue systems)
• modified approach to define radial mask on maps (only ppft and hollow procedures): search of outer radius starts from radius with minimum standard deviation

• restored option D to original interpolation and moved binning procedure to option H
• options 5 and 6: keep all info in first line of .dat files
• option 6: fixed bug when processing stacks with only one particle
• option j: when scaling, adjust also pixel size in .dat files

• figure of merit calculated in double precision

• avoid overflow error with metric/funcmode 0
• change metric 1 to normalized version
• use the same user-defined metric for origin refinement

• added option to apply the inverse of input mask

• input shifts now expected as pixels
• improved estimate of the background value
• write the map in a format compatible with p3dr

• estimate resolution at threshold value by linear interpolation
• added option -c/–cutoff to specify threshold value where to estimate the resolution

• changed installation procedure: auto3dem_install is the new command
• made easy to add support for new compilers

• maskmap: apply an input binary mask after adding a soft edge

• a minimum number of micrographs is used to generate an initial model. This value can be controlled by option -nmgmin in setup_rmc
• initial master file contains commented line for limiting number of cpus used at reconstruction

• provided binary for CTFFIND4 in the package (Linux only)
• at start programs will display if compiled in parallel or serial mode
• maps are flattened with a soft mask, using full map as reference for masking in gold standard
• inconsistency of pixel size in map header does not throw an error anymore

• added parameter to control the soft edge with solvent flattening (auto flatten_falloff)
• keep input/output files used by flatten and hollow procedure
• continue parameter file generated at end of each run has now unique name to avoid overwriting and to keep track of the processing

• option D modified to perform binning of micrographs instead of interpolation
• added experimental option 3x in autopp for using ctffind4

• added option for performing binning in batch mode (-bin)
• improved font appearance
• remove unused read/write menus
• added warning if can not connect to display
• minor fixes on size of file names
• avoid including origin when calibrating maps for subtraction

Implemented the gold standard approach, and set as default

• added installation script for Mac OSX
• added installation instructions for Mac OSX

• a3dm_plot_fsc.py: plot FSC curves from a given set of iterations
• a3dm_fsc2xml.py: convert FSC curve to EMDB-compatible XML format

• changed method to identify score to use for selection of particles
• setup_rmc now prepares parameter files for gold standard procedure
• map radial profile is avoided if new_ptles set
• move the parameter updates outside of even/odd loops
• make sure that parameter files are complete with particles for gold standard
• fixed bug when estimating mask radii and changed constant factor for outer radius

• added option 6 to split parameters into even and odd numbered particles
• option j: handle multistack parameter files
• option 6: check presence of 'dat' in parameter file names
• option 6: handle files not containing _ character when splitting into even/odd files
• option 6: make up particle parameters when splitting empty files
• options F and O: added option to delete input files once processed
• option 5: fixed split and join of parameter files
• option K: remove box size from first line of dat files
• added option for not masking corners when generating/processing stacks
• option j: remove box size info from first line of dat files
• option 3 (ctffind): remove unused variable pave

• increased output format size for pixel count

• handle large maps

• added warning messages during initialization
• fix error messages in get_info module
• constrained shift values in output dat files
• improved logs and check controls
• fixed index bug for compare_mode 1 (mode not currently used)

• constrained shift values in output dat files

• fixed interface and handle endianess

• accept unsigned integer format

• added -fastload to avoid calculation of radial profiles
• calculate radial profile also for maps loaded for comparison (Subtract 3D Map menu)
• redefined map_1d_rob as calc_map_1d_profile, and simplified its interface
• allow initialization of large arrays
• allow opening of large maps
• allow large indexes when reading integer maps
• changed window parameters to avoid colored stripes in the gui
• fix radial projection for large maps
• increased buffer display
• open stack images even if in different directory than dat file

• Changed license to be in agreement with University recommendations. Essentially the software remains free to use, modify and distribute for educational, research and non-profit purposes

• Added list of libraries required for installation in file INSTALL

• Fixed compiler warnings
• Updated CTFFIND3 to ver. 3.5, march 2014 snapshot
• Resolution range for estimating the defocus constrained to three times the pixel size, not to go beyond 6 angstroms
• Spherical aberration allowed to be zero
• Programs add size info to first line of dat files to avoid stack files to be read and checked at every run

• Fixed constraints on length of file names (4.04,4.03.1)
• Modified all metrics for parameters refinement to report the score as phase residual, in degrees
• Optimized some metrics (funcmode 0, 3, 4 and 5), and added the absolute value of the Contrast Transfer Function as a weight (4.04,4.03.1)
• Fixed variable casting and array indexing in order to process large maps and write them to file
• Changed default fonts for all visualization programs

• Fixed constraints on length of file names
• Fixed program to visualize large maps
• CTF analysis: display correct resolution frequencies when zooming on
• CTF analysis: allow flickering of all curve plots superposed to spectrum

• Fixed parsing of arguments
• Added option to split/join stack parameter files (4.03.1)
• Changed description message of option Z to reflect the ability to convert either map or stack files in PIF format to CCP4 format
• During ctf estimation, mrc files are deleted only if generated by autopp

• Avoid generating selection files for empty parameter files (4.03.1)
• Added option for automatically determining, at each iteration, the optimum radial range to use for masking the reference map (auto hollow_auto)
• Provide by default an optimum radial range for template masking at the end of the Random Model Computation
• Changed score selection approach to allow default values for parameters ‘auto score_fraction’ and ‘auto pft_cc_fraction’ to be present in the _master and _continue files
• Added option to automatically remove background density from the reference map (auto flatten_map)
• Changed behavior of ‘auto freeze_res’ in order to only freeze the range of resolution frequencies used for determining particle parameters, but to allow the reconstruction to be performed according to the resolution estimate
• Enforce a resolution-limited approach when automatically determining the resolution range used to estimate the particle parameters, in order to minimize noise overfitting
• Name of script files generated when only the reconstruction is requested during one iteration (‘auto generate_map’ set to ‘only’) are different than conventional ones, in order not to overwrite previous ones when stored
• In po2r, set default parameters to apply only-phase correction to images (ctfmode 2) and use the weighted normalized cross-correlation (funcmode 4) as metric
• Run ctf refinement as a separate iteration, precisely as the first iteration when the option ‘auto ctf_refine’ is activated
• Changed default ctf correction for ctf refinement to only-phase flipping
• Allow use of value zero for ‘filter’ option, corresponding to applying rather the correcting for the CTF
• Removed options filter, ctf_ff1 and ctf_ff2 from po2r and pctfr programs
• In pctfr, added options dfocus and nfocus to specify defocus range to search, anastigm to indicate if the CTF is to be assumed anastigmatic, and per_ptle_ctf to estimate the defocus separately for each particle
• In pctfr, changed the meaning of dangle to angular step of astigmatism angle

• Remove weighting parameter in reconstruction process, which was based on the standard deviation of the first image for each stack
• Particles with wrong information on their origin are discarded even if not previously flagged
• Fixed bugs that prevented reconstruction of large maps

• Optimized routines for parallelization and fixed bugs
• Added more parameters for better control of defocus and astigmatism angle range where to search
• Added ability to estimate defocus separately for each particle
• Removed internal constraint that prevented to search for astigmatism in stacks where the defocus estimate provided in the parameter file is anastigmatic
• Added option to enforce anastigmatism when refining the estimate

• Add constrain on origin coordinates

• Add new program flatten to automatically remove background density from a map
• Add support for large maps in zerodens
• Patch mrc2pif to accept mrc mode 6

Compilation

• Makefiles reorganized for better control of warnings
• Added make.inc.compflags where all flags are combined after being modified by single programs, if needed

General modifications

• Extended limits on number of micrographs and parameter files
• Extended length limit for names of mrc files
• Almost all the files have been analyzed and modified in order to minimize the warnings at compilation

Common libraries

• Fixed alignment of derived type pifHead. Problem seems specific to Intel compiler
• Changed many integer variables from 2-bytes to 4-bytes format

robem

• Fixed linear gradient correction when applied to data with negative values
• Fixed blemish removal bug
• Removed excessive output messages to terminal
• Avoided generation of cmap.robem file
• Changed default values for thinning and dilation in boxing gui and inverted their sequence of operation: first thinning and then dilation.
• Improved autobox routine by modifying some preprocessing steps and adding a post-selection screening based on screening tails of score histogram.
• Added standard deviation to output of spherical average when invoked by command line
• Fixed bugs when handling maps larger than 1024 pixels
• Some subroutines moved out of robem.c, including statistical analysis of images and autoboxing

autopp

• Added dilation parameter for autobox routine, option M
• Added parameter in option 4 to skip generation of dat files from ctf info if values are beyond given range
• Modified option 3 to delete temporary files generated for ctf estimation

auto3dem

• Added generate_map parameter to optionally split reconstruction and alignment into two separate steps. Accepted values are only (only reconstruction), yes (alignment and reconstruction), and no (only alignment)
• Added partrad parameter, which is dynamically updated to reflect the particle radius as estimated from the latest reconstruction
• Added adapt_angle flag to determine adaptively the angle step for search/refinement by ppft/po2r, using the particle radius and the current resolution level of the reconstruction
• Changed criterion to determine when refinement supersedes search mode. Now the switch happens when the improvement between last two iterations is less than 0.5 angstroms
• Changed default parameters for ctf amplitude correction in po2r and p3dr in order to reduce the contribution of low frequency components
• Changed definition of capsid score, used to select the best candidate map during the random model computation

P3DR

• Fixed determination of zero padding factor when requesting generation of smaller maps
• Fixed several routines to handle large maps (linear size ~>1300 pixels)
• Memory footprint optimized, especially for serial version
• Added optional precompilation flag, in Makefile, for calculating the map in single precision. Useful to reduce the amount of memory needed

POR

• Added substep refinement iteration around best orientation during local refinement
• Changed default weighting in the distance metric, which now accounts for the radial frequency

PSF

• Changed calculation of frequency bins in order to be equal size rather than equal number of contributing voxels
• Modified FSC calculation to properly account for phase differences

tiff2pif

• Fixed binning option for micrographs with short integer data format

Compilation

• Added compiler flag for large files support (4.01.12, 4.02)
• Updated flags for Intel compilers (4.01.3, 4.01.11)
• Minor cleanup and updates (4.01.8)

General modifications

• Minor cleanup throughout code base and update to F90-style (4.01.9)
• Modified all I/O routines in order to access to large files (> 2GB) (4.01.12)

Common libraries

• Code cleanup (4.01.9)

robem

• Code cleanup (4.01.6, 4.01.8, 4.01.9, 4.01.11, 4.02)
• Minor reorganization of source files (4.01.1, 4.01.9)
• Minor fixes for improving portability (4.01.2)
• Fixed bug in rotational filtering (4.01.6)
• Box coordinates are saved in data headers of boxed images. This makes it possible to delete particle images and generate new bcrd files that are consistent with the image file (4.01.5)
• Added simple approach in autoboxing routine, to filter X-ray-corrupted pixels from micrographs (4.02)
• Statistics for removing blemishes from boxed particles is now calculated on enclosed circle rather than on the full box (4.02)

autopp

• Fixed some typos (4.01.3)
• Modified options B (creating boxed image files) and K (recentering boxes, creating new bcrd and dat files) to better handle instances of multiple matches (4.01.5)
• Modified option E (batch conversion from tiff to pif format) to handle inversion of contrast (4.01.7)
• Modified option 2 (generating tiles from micrographs) to get directly tiles from micrographs instead of just their coordinates (4.02)
• Added option X to perform conversion from dm3 to pif format in batch mode (4.02)
• Added threshold argument to option M (autoboxing) for discarding low quality particles (4.02)
• Added argument to option 3 (ctf estimate) to output power spectrum from defocus estimation

auto3dem

• Exposed several refinement modes available in POR. The modes are selected by the new field ‘po2r mode’, and they can be: local (original local refinement), global(search over a grid extending the full asymmetric unit, requires new parameter glob_anglestep), mag (magnification refinement, requires new parameters nmagref, magref_step, magref_reset and magref_calibrate), ticos_equiv (search over the 60 symmetry-equivalent solutions). The change to interface is backward compatible with input parameter files from previous versions of auto3dem (4.02)
• Shell scripts made compatible with dash (default shell in Ubuntu flavor) (4.02)
• Minor bug fixes (4.01.11)

capsid_generator

• New program to generate PDB files listing coordinates of the points in an icosahedral lattice (4.01.10)

ctffiind3/ctftiltt

• Code refactorization (4.01.9)
• Fixed bug in ctffind3.f that was making the program hung because no tiles were selected (already fixed in the current Grigorieff's implementation).

ctffit

• Program removed. Deprecated in favor of ctffind3 (4.01.8)

dm32pif

• New program to convert micrographs from dm3 format to pif. A binning option is included (4.02)

emprj

• Fixed bug on CTF parameters (4.01.11)

imgstatst

• fixed nearest neighbor interpolation used for calculating radial profile (4.02)
• Fixed typos (4.01.11)

leginon_to_bcrd

• Fixed bug in setting the box radius (4.01.9)

mrc2pif

• Improved ability to convert mrc files. Only 3D maps and 2D stacks of particles are automatically forced to odd size (4.02)
• Box radius written in the output file header (4.01.11)

P3DR

• Fixed bug in routine realtocomplx (4.01.4)

PCTFR

• Fixed bug that caused crashes with particles not well centered (4.02)

pifmapmask

• Minor updates (4.01.9)

POR

• Optimized the magnification refinement (po2r mode mag in auto3dem). Computational time for refinement of magnification is now comparable (or smaller) to an iteration of local orientation refinement. New features: - refinement around a previous solution (through option magref_reset in auto3dem); - minimization of scaling of the output map, to use if the magnification is not intended for calibrating the data against a gold-standard (through the option magref_calibrate in auto3dem) (4.02)
• Improved convergence of global search. Reference score for current solution is now calculated with latest settings, to guarantees that resulting map is an actual refinement of the previous one. Also, if particle orientation is changed, a local refinement follows.
• Fixed bug in magnification refinement mode, so that figure of merit stored in parameter file is correct (4.02).

PPFT

• Fixed bug to better estimate particle origins under all conditions. The estimate was always off by one pixel when using bin=2 and the original box size was such that the center was an odd number (e.g. box size: 215 → origin: 107) (4.01.12)
• Added tests that avoid the program to interrupt without messages (4.02)
• Minor bug fix (4.01.12)

tiff2piff

• Added 4th argument to command line to invert contrast (4.01.7)
• Code cleanup in I/O routines (4.01.9)

autopp

• Complete overhaul of program to make use more intuitive
• All output files generated in directory where autopp is run
• Input parameters can be provided at the prompts or specified on the autopp command line. If partial input is given, user will be prompted for remaining arguments.
• autopp command line written out if input provided at the prompt
• Option 4 (generation of particle parameter files from ctffind3 output) generates warnings if defocus levels or degree of astigmatism falls outside of user specified bound.
• Particle reboxing greatly simplified and provides two options for placement of new box coordinate and particle parameter files.

PPFT

• Bug fixes implemented so that program does not crash if unable to determine particle origins. Results are still meaningless for these particles, but segmentation faults are avoided.

robem

• Bug fix so that defocus major and minor values are not assigned values before pixel size has been set
• Radial in radial projection window now shown in both pixels and Angstroms

ctffind3

• Code optimized to run up to 3-3.5x faster than previous version under certain conditions (typically smaller micrographs, larger defocus search range, and smaller defocus search step size)

pifmapmask

• Replacement for pifmapdel that provides options both for deleting or masking volume covered by PDB structure.

parse_dm3

• Note added to output alerting users that certain hardcoded parameters are specific to Baker lab microscope calibration

fixpif

• General cleanup to simplify code and improve maintainability

NOTE – Following change had been implemented in v3.15, but was not described in previous version of this document:

Simplified the process of selecting CTF nodes in robem 'FFT – CTF Estimation' window. By default, images are assumed to be stigmated, but user can toggle between choosing parameters for stigmated or astigmatic images.

Building executables

• The make_all script unsets the LANG environment variable before building executables. In some cases the default value for LANG can interfere with the build, particularly for robem and other GUI applications.

auto3dem

• Improvements made to program imgstats to provide additional statistics on either PIF maps or the individual images in a boxed image file. Program has also been integrated into auto3dem to write out map statistics after all iterations.
• No longer need to specify 'ppft verbose -1' in the auto3dem master file in order to avoid expensive calculations of 3rd correlation coefficient (cmp_cc). Program PPFT has been modified so that verbose = -1 is the default.
• Automatic determination of capsid inner and outer radii is now much more reliable and faster to compute. Radii are determined directly from radial density profiles and no longer rely on the cmp_cc coefficient from PPFT.
• Parallel jobs can now be launched by auto3dem using ether mpirun or aprun, thereby providing greater flexibility regarding parallel computing environment. At run time auto3dem checks for the existence of these programs and automatically chooses the appropriate one.
• Negative values in FSC curve set to zero before attempting to apply noise suppression algorithm in P3DR. Avoids generation of pathological map where all pixels have the same value.
• Source code (include/allprogs.inc) updated so that maximum number of CPUs is now 512.

Defocus estimation using ctffind3

• The ctffind3 and ctftilt programs (Grigorieff lab) have been integrated into the auto3dem software. The source code is located in the ctfniko directory and the executables are built at the same time as other auto3dem applications when the make_all script is run. These executables have been named ctftilt_a3dm_mp and ctffind3_a3dm_mp to avoid conflicts with existing installations. Defocus estimation can be done on multiple PIF or MRC micrographs using autopp Option 3 and particle parameter files generated using Option 4.

setup_rmc

• Capsid radii are estimated from the density profiles of the maps and automatically updated after all iterations. This greatly improves the likelihood that an individual random model computation (RMC) will converge to the correct low-resolution structure.
• The auto3dem master file generated by setup_rmc sets the default number of iterations to 10, no longer generates the commented out 'ppft verbose -1' line, and comments out the 'auto bin_reduce 1' line. The latter change was made since running PPFT with un-binned data before transitioning to auto3dem refine mode typically does not lead to higher resolutions.
• The algorithm used to select the best model from multiple models has been made more reliable. It was found that FSC-based methods did not always choose the best model, particularly for smooth, featureless capsids.
• The original RMC can be selected with the –trad flag.

robem

• Major overhaul of code to avoid use of User Interface Language (UIL) files for the specification of initial state of GUI. This greatly improves code readability, portability, and maintainability.
• Isolation of many GUI operations so that robem can be run in batch mode (normally through control of autopp) without having to launch Xvfb. Before making these changes, getting robem to run in batch mode was very problematic and often the biggest hurdle to portability. No changes are necessary when using autopp since modifications to robem syntax have been implemented. When running robem standalone in batch mode, need to make following change to calling sequence

• Fixed the feature that adds Gaussian noise to an image. This had been done incorrectly in previous versions as evidenced by the fact that adding Gaussian noise with mean=0 changed the mean intensity of the image. (This feature is found at Manipulate > Fix Values)
• Fixed bug that caused robem to crash when clicking in drawing area of particle processing screen before particles had been loaded.
• Fixed bug where display of radial projections would sometimes cause robem to crash.
• Modified code so that error messages no longer generated when performing zooming on a micrograph.
• Overhauled 'Point Value' screen so that data is easier to interpret, distances are provided in Angstroms if the pixel size is listed in the map/image header, and behavior of mouse buttons is consistent with behavior that most users would expect.

gensph

• Program for generating a spherically averaged PIF map from an input map

pifmapdel

• Program for masking a PIF map using PDB file coordinates

prjcompare

• Program for calculating the difference between two image stacks. Typically used to compare boxed images to projections of a map.

parse_dm3.pl

• Script for parsing DM3 file header and extracting voltage, spherical aberration, magnification, and pixel size. The latter quantity is specific to the Baker lab and the get_pixel_size function will have to be modified for other microscopes.

PIF library and calls to PIF library

• Initialize the character string used to specify the read/write arguments that are passed to pif_open. On some machines, approximately 1-in-1000 calls to pif_open from P3DR randomly set the second character in the string to a value that caused the program to crash.
• General cleanup of PIF library, including standardization of commenting and elimination of obsolete code related to storage of binned data on disk.
• Obsolete functions or functions with redundant capabilities were eliminated. Applications that use the PIF library were modified accordingly. The list below contains the deleted functions, with the replacement calls, if needed, in parenthesis:
  1. pif_read_tmapi2 (pif_read_mapi2)
  2. pif_read_imgi2 (pif_read_short_image)
  3. pif_read_imgi1
  4. pif_read_byteshort_image

auto3dem

• Bug fix to P3DR so that it can handle case where the first particle parameter file is empty. Behavior when subsequent parameter files are empty is unchanged.
• Modifications to setup_rmc to write scripts that simultaneously launch and manage multiple random model calculations under control of Portable Batch Scheduler (PBS). All currently running jobs are terminated and scheduled jobs de-queued if any calculation terminates early upon finding a suitable starting model.
• Add new keywords to restrict particles used in reconstruction on the basis of orientation angle omega. Select all values within omega1_tol of omega1 OR omega2_tol of omega2, while properly accounting for behavior at 0°/360°. By default, all values of omega are selected.

robem

• New buttons added to FFT dialogue so that user can choose between writing full (header + particle records) or empty (header only) particle parameter files.
• When writing out boxed image and box coordinate files, default names automatically provided. For example, if the name of the micrograph being processed was micro.pif, the default image and coordinate files would be micro_box.pif and micro.bcrd, respectively.

imgstats

• New program for calculating statistics on maps and boxed image stacks. Provides global statistics and radial profile information. Note that output for image stacks can be quite large since a radial profile will be generated for each image.

Code cleanup

• In addition to modifications listed above, numerous changes were made to the code to improve portability and maintainability. Debug code and old code that had been retained for compatibility with pre-ANSI C was removed. Unused functions were deleted and redundant routines were merged into single versions. Obsolete or unnecessary variables were removed from argument lists and required changes to calling routines were made. A number of files that were no longer needed after software cleanup were removed: symlib.c, imsubs2.f, parser.f, libR/appodize.c, emicolib.f, libCtf.c, vax_minmax.inc, em.cmm, pftlib.f, movc3.f.

• Note to developers: Major cleanup of the PIF library was initiated. More than 2000 lines of code were eliminated and redundant routines with very similar functionality were merged. All codes distributed with v3.15 have been modified to use the new PIF library, but old applications will either need to continue using the earlier version or be changed to handle the new interface.

auto3dem

• Feature added to POR to test the 60 orientations related to the starting orientation by icosahedral symmetry. Useful when trying to identify lower symmetry components of an otherwise icosahedral structure. Enabled as shown below

por ticos_equiv 1       # Add to auto3dem master file
ticos_equiv 1           # Add to POR input file

• Feature added to POR to perform global search of orientation space. Useful when trying to ensure that all image orientations have settled into the correct part of the asymmetric unit. Use with caution since this option can VERY expensive since it involves a 3D search over (q, j, w). Enabled as shown below

por global_por 1       # Add to auto3dem master file
global_por 1           # Add to POR input file

• Added tests to POR and PPFT to make sure that input map is present and readable. Program terminates and error message written if readable map not found.
• P3DR now writes out the number of empty voxels at each spatial frequency. This output can be used as a diagnostic to determine if transform of map is being too sparsely populated at a particular spatial frequency.

robem

• General improvements to robem including:
  • Setting of sensible defaults (e.g. 1D defocus estimation rather than 2D)
  • More meaningful labels provided
  • Widgets resized and repositioned to fit properly on screen
  • Clearer behavior of toggle buttons

• Code modified so that bulged icosahedral projections are displayed properly. For large maps, would sometimes see artifacts at equator.
• Partial overhaul of code to improve readability and maintainability.

oned, ctfdisp, fixpif

• No longer need to set EMDIR environment variable in order for programs to find corresponding UID configuration file. File now found by either searching PATH variable or determining location where program is installed when full path is specified to executable. General cleanup of code and elimination of unused routines.

emprj

• Eliminated discrepancy between code and documentation for CTF parameter list. Cc, beta, and dE must now be provided.
• Fixed bug so that projections of map are now multiplied by CTF. Previously emprj would fail to apply any correction when one of the 'multiply by CTF' options was chosen.

New PDB manipulation scripts

• Although these are not strictly related to image reconstruction or analysis, they may be useful for performing basic tasks related to constructing starting models from PDB coordinates.
  • pdbcalpha – retain only Ca atoms.
  • pdbclean – retain only TER, REMARK, and ATOM records
  • pdbcomb – combine two PDB files
  • pdbscale – scale atom coordinates

New box file format conversion scripts

• The EMAN boxer program and robem produce box coordinate files in different formats. To enable easy conversion between the two, the following scripts have been created
  • box2bcrd – convert EMAN box format to robem bcrd format
  • bcrd2box – convert robem bcrd format to EMAN box format

NOTE – As of version 3.13, when the software suite is built using the make_all script, the symbolic link BIN pointing to the bin directory is no longer created. In your configuration file, make sure that your path is set appropriately.

Also (see robem below), the EMDIR environment variable no longer needs to be set. It is now sufficient just to set the path to point to the location of the executables.

auto3dem

• Backup of summary file automatically created if launching a new, as opposed to a continuation, run. Avoids accidental overwrite of results summary.
• Format of summary file modified to include number of micrographs and range of defocus values used in the reconstruction.
• More robust error detection and reporting. Error messages provided showing parallel program and input file being used when abnormal termination occurred.
• Random model computations modified to handle concatenated particle parameter files and image stacks in MRC format. When launching setup_rmc, data directory automatically searched for parameter files with suffixes as large as 999.
• Output routine called by P3DR modified to avoid MPI buffer overflows and allow writing of maps of size larger than 10243.
• Cleaner handling of symmetries other than icosahedral (532). Specification of symmetry in auto3dem master file using 'auto symm_code n' overrides symmetry parameters set for all other programs.
• Control of handedness tests in POR now under control of input parameter. Can be set either in auto3dem master file using 'por handtest [0|1]' or POR input file with 'HAND [0|1]'. By default handedness test is still performed.
• Magnification factor refinement capabilities added to POR, but not yet integrated into auto3dem. Controlled using 'MAGR nmagf magfstep' in POR input file, where nmagf is number of steps to take in each direction from magnification factor equal to 1.0 and magfstep is the step size.
• Particle images with centers that are too close to edges of box (< 10% box width) are excluded from origin and orientation refinement in POR. This is done to prevent unexpected termination of program.
• Flush operations added to P3DR to force writes of buffered data. This makes it easier to track progress of reconstruction and identify input images that cause P3DR to crash.

robem

• No longer need to set EMDIR environment variable in order for robem to find corresponding UID configuration file. File now found by either searching PATH variable or determining location of robem when full path is specified to executable.
• Difference map now printed along with 'second' map from robem difference map window.
• Modification to 'distance' feature in Point window so that distances can be measured when working with image stacks.

ctfdisp

• Display fixed so that resolution and value of CTF can be easily read.

In addition to changes listed above, extensive cleanup of underlying code: removal of obsolete functions; deletion of unused variables, common blocks, and macros; addition of INTENT specifications in Fortran subroutines; placement of frequently used sets of operations into functions; general cleanup to improve readability of code; simplification of I/O for maps and image data; modifications to enforce compatibility of C header and Fortran include files.

auto3dem

• Automation script and all programs that read image data (PO2R, P3DR, PPFT, PCTFR) modified to handle both MRC and PIF image stacks. Note that the input maps must still be in the PIF format and all output will also be generated in PIF format. This is the first step towards making the software completely compatible with MRC data.
• Particle images more than 50% of the box radius from center of box automatically rejected from reconstruction. Default can be overridden by specifying a value for the 'auto box_center_offset' parameter.
• Files listing the images that had been rejected from the reconstruction (e.g. as a consequence of particle origin being too far from center of box) are now generated and stored in the project’s dat/virus_INTFILES directory.
• Programs PO2R, P3DR, PPFT, and PCTFR now detect and automatically reject particle images that have invalid image numbers (negative, zero, or greater than the number of images in the corresponding image files). The presence of invalid image numbers formerly caused these programs to crash, but now simply results in a warning message in the output file.
• Image files for which the corresponding particle parameter files (.dat files) are empty are now properly handled by program P3DR. Although these micrographs never contributed to the reconstruction, they could occasionally cause P3DR to crash. Note that if ALL parameter files are empty, it is still assumed that ALL image data will be used.
• Tests on the microscope accelerating voltage and amplitude fraction made less stringent. For reconstructions from negative stain image data, the amplitude contrast is generally much larger than that for cryo and valid data sets were sometimes rejected by auto3dem.
• PO2R modified to handle image magnifications greater than one. In previous versions of code, higher image magnifications could occasionally cause segmentation faults.
• Summary file now lists the total number of particle images that could have potentially been used in the reconstruction.

autopp

• Particle reboxing (option K) now allows the user to specify the current working directory as the location of the particle parameter (dat) and/or coordinate (bcrd) files. New files would be written to the dat_new and/or bcrd_new directories.

setup_rmc

• Command line options added to control number of bins and resolution ranges used when calculating FSC curve.

robem

• Cursor used for boxing particles has been changed from 'cross' to 'diamond cross' shape, making it easier to see and track the cursor.
• Modifications made to repair bug that affected deletion of images from file.
• Text color changed on CTF estimation page to improve readability.

In addition to changes listed above, general cleanup of code base and removal of obsolete or redundant subprograms. make_all and clean_all scripts modified to use /bin/bash rather than /bin/sh since some flavors of Linux (e.g. Ubuntu) no longer support /bin/sh.

auto3dem

• Program P3DR modified to handle maps of size larger than ~12003. Similar modifications made to PO2R, PCTFR, and PSF to allow calculations on larger maps. Map size limit in the latter set of programs was estimated to be ~17003.

autopp

• Improvements made to option K (generation of new box coordinates from particle origins and original box coordinates) to provide greater flexibility in naming convention of bcrd files relative to dat files.

robem

• Bug fix so that 'Find mag' feature in difference map window works correctly.
• Modifications to automatically shift boxes that overlap edge of micrograph to lie entirely within the micrograph. Also made minor changes to code so that boxes very close to edges and corners of micrograph are handled correctly.

combine_fsc

• New program that combines the results from multiple Xmgrace .agr files to create a plot containing multiple FSC curves

Code cleanup

• Eliminate unused functions, files, and variables
• Remove redundant files
• Simplify logic of setting labels, titles, and values in robem widgets
• Replace non-standard function calls to float() with Fortran 90 real( )

auto3dem

• Produce FSC curves using Xmgrace instead of the GD package. PNG file is generated only if Xmgrace is found in the Linux/UNIX path, but project file is always written to allow manipulation or plotting of FSC curves after completion of calculations. Using Xmgrace rather than GD circumvents the known problem that GD has in plotting data with numerical (non-evenly spaced) x-values. Modifications were made to the auto3dem.pl file and the new modules/grace_utils.pm file was added.
• Modify logic so that FSC curve is only calculated if the ‘auto estimate_res’ option is set and move the creation of the filtered files inside the block of code used to create full map.
• Fix bug in libCommpk/ctf_para.f so that astigmatism is treated correctly. Had been making call to atan2(x,y) instead of atan2(y,x).
• Modify pftsearch.F and global_cc.F so that we either calculate projections of model using the origins listed in the .dat files or, in the case where all .dat files contain only header information, the center of the image box. Get rid of PPFT files map_sym_cavg_ppft.f and prjavtg_fft.f since these are no longer needed and update Makefile accordingly.
• Modify algorithm used in PPFT to minimize memory usage during generation of projections of the model. Required addition of new subroutine map_prj_slice.

autopp

• Option F (blemish and linear gradient removal, normalization) modified to test that all boxed image files in the argument list are readable.
• Option 3 added to allow batch processing of micrographs with ctftilt.
• per_ptle_ctf
• Minor bug fix so that –mode astig works properly

robem

• Modification and cleanup of writeBoxCoords functions so that bcrd files are written out immediately rather than waiting until completion of session.
• Default particle parameter files now have .dat_000 rather than .dat_001 suffix. This change makes it easier to correlate parameter files with auto3dem iteration numbers.

auto3dem

• Number of CPUs can now be specified using both -ncpu and -np flags
• Tests added to ensure that the inner and outer radii specified for programs PCUT and PPFT are both properly ordered (e.g. inner radius < outer_radius) and less than radius of boxed images
• max_cpu flag parameter now written to auto3dem restart and continue files for programs PO2R, P3DR, PPFT, and PCTFR
• Bug fix in PPFT to allow computations on maps larger than 8003
• Bug fix in PO2R to prevent out-of-bounds array access when using particle magnifications that are greater than 1.0

setup_rmc

• New command line option to specify maximum resolution of computed map
• New command line option to set radii used by programs PPFT and PCUT

autopp

• Generation of new boxed coordinate (bcrd) files from particle parameter files and old bcrd files now more robust with regards to file formats and naming conventions
• Bug fix to option G (auto-boxing, blemish and linear gradient removal, normalization) to avoid entering infinite loop

per_ptle_ctf

• Added capability to handle missing or out-of-order records in particle parameter files

robem and ctfdisp

• Corrected spherical aberration of objective lens for FEI Polara and Sphera microscopes from 2.0 mm to 2.3 mm and 2.26 mm, respectively. Previously solved structures should still be valid since incorrect values had negligible impact on CTF correction.
• Modifications to allow calls to Fortran from C using name mangling conventions employed by PathScale compilers
• General cleanup of code to consolidate functions that handle behavior of arrow widgets
• Collection of macro definitions into single location to avoid accidental specification of conflicting definitions

• I/O routines modified to handle embedded blank lines in particle parameter files
• Functions parse_key_input( ) and list_key_input( ) temporarily added back into libUtil. Needed in anticipation of possibly adding serial Fourier-Bessel image reconstruction program EM3DR to software release.
• Program terminates and error message written if attempting to use the following options together. Setting PPFT verbose flag to -1 disables calculation of quantities needed to calculate inner and outer radii of annulus

       auto freeze_annulus 0
       ppft verbose -1

Made changes to the source code that allows all programs except robem, ctfdisp, and emprj to be built using the PathScale compilers.

• Variable types changed from integer*2 to integer*4 so that arguments to mod() intrinsic function are of the same type.
• Parenthesis added around negated terms so that expressions like x = y + (-z) are parsed correctly.
• Created double-underscored versions of functions in PIF library to allow for multiple name mangling conventions. For example, func( ) func_( ) func_ _( ).
• Explicitly added #include <stdlib.h> to C source files where needed.

AUTO3DEM and all programs that are called by it (P3DR, PCTFR, POR, and PPFT) can now handle particle parameters files containing data for multiple boxed image files. These concatenated files are of the following form

image_file_name
defocus information
zero or more particle records
image_file_name
defocus information
zero or more particle records
…

No embedded blank lines are allowed in the concatenated particle parameter file, but trailing blank lines are still permitted.

Note that the script setup_rmc.pl, which is used to setup the random model calculations, still requires that each particle parameter file contain data for only a single micrograph. This may be remedied in future releases.

Other changes in this release include:

• All long lines (>72 characters) in Fortran source code have been shortened to fit within the standard fixed-format limit. This affected a small number of lines that were still syntactically correct after being truncated. The only effect of these lines was on the printing of rarely encountered error messages and in the display of the 'signal minus background' curve in the RobEM CTF estimation screen.
• Warnings are no longer produced when defocus values are outside of the range 0.8-4.0μ
• Deleted the return statement in pif2ccp4 main program. This had been causing errors for some newer versions of compilers.
• Added xbatch.mscp to bin/ directory to allow RobEM to be run in the background in batch mode.
• Addition of new options to autopp: renaming of file prefixes and generation of BCRD file for tiling scanned image file.
• Make script now moves (instead of copying) all executables into bin/ directory.
• General cleanup of code for programs fixpif, ctfdisp, and oned.

• Modified image reconstruction programs to handle up to 2048 particle parameter files.
• Bug fix to programs mrc2pif and pif2mrc to allow proper conversion of image data in BYTE format.
• Bug fix to RobEM map projection feature in general and generation of projections spanning the icosahedral asymmetric unit in particular. Delete routines from RobEM source and replace with calls to libCommpk.
• Bug fix to 3D Display panel in RobEM that caused min/max radius data to be displayed in the wrong text box. This bug was introduced in v3.05 and had no effect on RobEM results.
• Increased width of text box that lists file names in Circular Average display. Previous box width was too narrow to allow longer file names to be read.
• Clicking on section button on main tool bar now displays central section of map rather than first section.
• Point screen panel now contains button for identifying and drawing marker at the center of a section.
• Improvements to handling of early termination mechanism in random model calculations
  • RMC_DONE file automatically deleted during the random model calculations.
  • Existence of the RMC_DONE file no longer triggers termination of auto3dem. This logic has been moved into the auto-generated RMC_run script.
  • Message is written both to stdout and the auto3dem log file informing the user that the 'auto quit_early' flag had been set and that this feature is normally used only during the random model calculations.

• Modified communications in P3DR (exch_intp.F) and PPFT (pftsearch.F) to avoid MPI buffer overflows when working with large maps.
• All Fortran source code updated so that lines are less than or equal to 72 characters. Compiler flags used for extended Fortran source have been deleted from Makefiles and options added for PathScale compilers.
• Declaration of maximum number of particle parameter files and length of filenames now declared in a single include file (sizes.inc) that is used throughout all image reconstruction codes.
• Arguments for SYSTEM_CLOCK function changed from integer*8 to default integer.
• PIF library modified to write out parameters a0-a5 from CTF estimation procedure to global file header.
• Added capabilities for interpolation of packed boxed images to RobEM.
• Major cleanup of RobEM source code: improved formatting of loops and blocks; removal of non-informative comments; standardization of function layouts; removal of common macros/definitions in UIL and header files, placement into common definitions file, and modification of Makefile to dynamically generate content; deletion of unused or obsolete functions; combination of all widget deactivation functions into a single generic function.

A number of important new features have been implemented in this release.

• CTF corrections can now be applied on a per-particle image basis in programs PPFT, PO2R, and P3DR. Three steps are needed to use this feature. Note that the program CTFTILT is not maintained by our laboratory and must be downloaded from http://emlab.rose2.brandeis.edu/grigorieff/download_ctf.html
  1. Run CTFTILT program on micrographs.
  2. Generate extended particle parameter files from CTFTILT output file and standard format particle parameter files using per_ptle_ctf script
  3. Add the following line to the auto3dem input file, where auto enables the feature for all three programs

(auto|ppft|po2r|p3dr) per_ptle_ctf 1

• Noise suppression algorithm described in Rosenthal and Henderson, JMB 333 721-745 (2003) is now available in P3DR. Add following line to auto3dem input file

auto noise_suppression 1

• Program PPFT can now be run using all empty (except for header) particle parameter files. Number of particle images and box center are obtained from corresponding boxed image files specified in particle parameter files.

• Programs PPFT, PO2R, PCTFR, and P3DR now allow embedded comments in their input files. Everything after the first ‘#’ character is ignored. Previous versions only allowed comments where ‘#’ was the first non-whitespace character. Note that this feature only impacts users who run these programs standalone (not through auto3dem).

• Improved auto boxing of empty particles in RobEM, bug fix when writing movies with .pcx file naming, and enhancement of help facility.

• Variables specifying number of views per read deleted from all source code. No longer needed since overhaul of parallel algorithms in auto3dem v3.03. Corresponding parameters now ignored and marked as deprecated in auto3dem.

Other minor modifications and bug fixes in this release

• Cleanup of #include statements in RobEM source code
• Overhaul of UIDATE subroutine in unix.f to improve portability
• Rearrangement of order of elements in user defined type in strucfac.inc for better alignment of variables
• Explicit casting of variable in tiff2pif.c to avoid compiler warnings
• Fix error in specification of argument intents in pftcc_peak.f
• Minor bug fix related to error handling during terminal input in sflib.f
• Change MIN0 and MAX0 function calls to MIN and MAX respectively in misclib.f
• Minor changes to make_all and clean_all
• Deletion of robem.uil – superseded by new_robem.uil
• Deletion of libCompar/read_files.F and libCommpk/readorient.f since they are no longer needed after completing cleanup of I/O
• Replacement of get_ptle_params.f with get_ptle_ioom.f and get_ptle_ioomctf.f to allow more flexibility in handling per-micrograph and per-particle CTF corrections
• Renamed map_prj_ppft.f to map_prj.f and moved to libCommpk; deleted routines map_prj, map_prj_axis, map_prj_xz, and map_prj_all from maplib.f; modified pftsearch.F to call map_prj instead of map_prj_ppft; modified emprj.f and emprj_xdy.f to use call new version of map_prj.

• Fixed bug that affected particle selection from PPFT-generated parameter files.
• Parameters added for new microscopes in robem.
• Minor modifications to avoid compiler errors from newer versions of gcc/gfortran.
• quick_omega (PPFT) and quick_search (PO2R) now turned on by default. These changes only impact users who run these as standalone programs since the options had been automatically enabled when running through AUTO3DEM.
• Improved commenting and elimination of unused code in robem, oned, and ctfdisp.
• Simplification of correlation coefficient calculations in PPFT. Changes do not affect code performance or results, but greatly improve readability and maintainability.

• Overhauled parallelization scheme for programs PPFT, PO2R, and P3DR. Reduced memory requirements in PPFT by approximately 40%. Test problems run on 16 processors showing reductions in run time of 25-40% relative to v3.01.
• Utility program autopp added for performing repetitive tasks such as converting files from one format to another or globally replacing a string in a set of files. Note that not all are currently operational. This script should still be considered as under development.
• Fixed bug in PO2R that resulted in ctfmode being hard coded to 1.
• Defocus values in particle parameter files only tested if CTF corrections are used.
• Header always written to auto3dem summary file if file does not already exist, regardless of value of restart flag.
• Minor bug fix to PO2R so that parameter nangle specifying number of steps to be taken along each orientation angle can be set to zero. Doing this allows PO2R to refine particle image origins while leaving orientations unchanged.
• Modify tiff2pif conversion program to optionally accept third command line argument that specifies binning factor. Formerly, users were prompted to enter value.
• Add mrc2pif, pif2mrc, and pif2ccp4 conversion programs. Modify so that programs can either accept file names on the command line or prompt user to enter names.
• Provide options for building applications using static linking.
• Executables now reside in bin/ directory rather than BIN/. To ensure back compatibility, BIN is now a symbolic link that points to bin/.
• Improved instructions provided on using 'ppft verbose -1' option in auto3dem input file that is automatically generated by setup_rmc.
• Move perl modules (.pm files) into modules/ directory and modify perl scripts to look for modules in new location. Replace 'use lib do' construct with 'use FindBin'.
• Add documentations files that are used by programs robem, oned, ctfDisp, and fixpif.
• Minor cleanup of argument list for subroutine global_cc in program PPFT. Also affects calling routine global.

• Minor modifications to make files to ensure that applications build properly when using Portland Group compilers. The –Mnomain option is now used only with those applications that require access to Fortran libraries, but have main program declared in C source. Remove getarg subroutine from libCommpk to avoid clashes with Portland Group library function. Declare iargc to be of type integer in routine parse_cmd_line.
• Add mrc2pif and pif2mrc conversion routines to convert directory.
• Modify calls to XtVaSetValues() from function defocusArrowActivate() in ctfDisp program to avoid segmentation faults on 64-bit machines.
• Cleanup preprocessor directives in robem source.

This release gathers for the first time all of the image reconstruction, image preprocessing, and utility codes into a single archive. New additions include robem, ctfDisp, oned, fixpif, tiff2pif, emmap3dt, emmapzoom, emsf, em3dbt, emprj, diffit, normit, zerodens, and a number of conversion routines. The make_all script now takes two command line arguments

make_all (parallel | serial) (all | gui | nogui)

The first argument behaves exactly as before, while the second argument specifies which set of applications should be built. In most cases, the second argument should be ‘all’, but the ‘nogui’ option may be chosen when doing a build on a machine that does not have the Motif library installed. The ‘gui’ option is generally not needed, but is provided for added flexibility.

If you are already a user of robem, please be aware that this release contains a major overhaul of the software. Modifications include:

• All Fortran code has been brought up to the Fortran90 standard
• Much redundant, obsolete, or unused code has been deleted
• Old Kernighan and Richie (K&R) style function prototypes have been replaced with prototypes that conform to the ANSI C standard
• Directory structure, make files, and include files have been cleaned up and reorganized
• Minor bugs, mostly related to type mismatches in C function calls, have been implemented
• Routines used by multiple applications have been moved into libraries

In order to build many of the newly added programs, you will need to install the Motif library. We do not currently have a feature in place for automatically determining whether the code is being compiled on a 32-bit or 64-bit architecture. The make.inc.common file is hardcoded for 64-bit hardware and the MOTIFLIB macro will need to be manually edited if you are running on a 32-bit machine.

Running the new programs requires that the EMDIR environment variable be set. The easiest way to do this is to set the EMDIR variable in the .cshrc file and then use it to append the path variable.

setenv EMDIR /path_to_programs/BIN
set path  = ($EMDIR $path)

All make files have been overhauled and a number of subroutines that are common to both the program PPFT (parallel PFTsearch) have been moved into libCommpk. In addition, several of the routines in the PPFT directory have been renamed so as to avoid conflicts with other routines of the same name in libEMF.

fft_map_fill           renamed         fft_map_fill_ppft
map_fft_fill           renamed         map_fft_fill_ppft
map_prj                renamed         map_prj_pfft
map_sym_cavg           renamed         map_sym_cavg_pfft

The following directories have been created to accommodate the newly added code: convert, conway_tif2pif, ctfDisp, em_tools, fixpif, lib3DAll, libEMF, oned, pftprj, robem.

• Overhaul make files and define library macros in include files. Rename libraries to use more standard naming conventions and rename directories as follows to be more consistent with other Purdue software

          PIFlib         renamed         libR
          Commpk         renamed         libCommpk
          Compap         renamed         libCompar
          DIElib         renamed         libDIERCKX
          Vfftpk         renamed         libVfftpk

• Add new script flesh_out.pl to generate complete particle parameter files from files that contain only header lines.
• Fix minor bug in select.pm related to filtering particles on theta and phi
• Get rid of old code delimited by #define OLDWAY preprocessor directives in libPIF.c
• Generate new libPIF.h header file from libPIF.c

Important notes for building auto3dem

Auto3dem and the parallel codes that it calls can now be built and run in either serial or parallel mode. An implementation of the MPI library (e.g. mpich) is no longer needed when running auto3dem on a single processor. The make_all script used to build the executables now takes a single command line argument, with allowed values of ‘parallel’ or ‘serial’. Symbolic links are set by the script as follows

% make_all parallel
% ls -g make.inc BIN/mode.pm
lrwxrwxrwx  1 csd357 14 Jun 13 15:14 BIN/mode.pm -> mode_parallel.pm
lrwxrwxrwx  1 csd357 15 Jun 13 15:14 make.inc -> make.inc.parallel

% make_all serial
% ls -g make.inc BIN/mode.pm
lrwxrwxrwx  1 csd357 14 Jun 13 15:14 BIN/mode.pm -> mode_serial.pm
lrwxrwxrwx  1 csd357 15 Jun 13 15:14 make.inc -> make.inc.serial

When building auto3dem for parallel operation, the serial Fortran 90 and ANSI C compilers are called by the corresponding mpif90 and mpicc scripts. For builds in serial mode, the FC and CC macros in the make.inc.serial file must be manually edited if you are not using the gfortran and gcc compilers. The make_all script determines whether or not a build of auto3dem already exists. If the mode (serial or parallel) of the current build differs from that of the previous build, all .F files are ‘touched’ to ensure that they are run through the preprocessor and recompiled.

For auto3dem runs in serial mode, the number of CPUs no longer needs to be specified and any value set using the –ncpu flag is ignored.

Detailed listing of code changes:

• auto3dem.pl and run_mpi_prog.pm modified to use mode.pm module. When running in serial mode, auto3dem.pl no longer requires (and quietly ignores) specification of number of nodes. validate.pm module modified to accept mode argument. Version string now specifies whether serial or parallel build is being used.
• make_all script modified to accept mode (serial or parallel) on command line and set symbolic links to appropriate make include files and Perl modules before initiating build. Make include files simplified to just make.inc.parallel and make.inc.serial.
• Argument intents added to P3DR/exchange_2_slab.f
• File extension for P3DR/realtocomplx.F, matrixentries_slab.F changed to .f

• Makefiles for POR, PCTFR, PSF, PPFT, and PCUT modified to include generic rule for creating objects from Fortran source with .F extension.

• Add preprocessor directives to following files to isolate parallel code and allow optional builds in either serial or parallel mode. Add serial code where needed to replace functionality carried out by MPI routines (e.g. initialization of process number and number of nodes, data reduction). All files with .f extension renamed to use .F extension so that they will be recognized by preprocessor.
  • Compar/bcast_parameters.f
  • Compar/error_stop.f
  • Compar/exch_3d_1.f
  • Compar/gather3d.f
  • Compar/output_density.f
  • Compar/read_files.f
  • Compar/read_map.f
  • P3DR/P3dr.F
  • P3DR/cmpt_intrps.F
  • P3DR/exch_intp.F
  • P3DR/exchange_2_slab.f
  • PCTFR/Pctfr.f
  • PCTFR/cmpt_ctf.f
  • PCUT/Pcut.f
  • PCUT/cut_map.f
  • POR/Por.f
  • POR/cmpt_ort.f
  • PPFT/global.f
  • PPFT/global_cc.f
  • PPFT/pftsearch.f
  • PSF/Psf.f
  • PSF/comp_sfactor.f

• Cleanup and modernization of routines in PPFT
  • Makefile – remove map_peak_rest.o from object list
  • map_peak_restr.f – remove; functionality now in pftcc_peak
  • list_ccs.f - cleanup and intent specification
  • pftcc_peak.f - major cleanup and reorganization for clarity; move call to rearrange_fft_style from here to get_xy; replace call to map_peak_rest with inlined functionality.
  • get_xy.f - add call to rearrange_fft_style after call to ccf_fft
  • calc_pfts_g - add FILT_FAC to argument list; declare pfft to be complex rather than real and make necessary code changes
  • tsbend.f - general cleanup and intent specification
  • rearrange_fft_style.f - rewritten so that rearrangement of array done in place. Single array now passed with intent INOUT and same dimensions as array in calling routine.
  • pftsearch.f - get rid of variables NPHI and NTHE
  • global.f - get rid of variables NPHI and NTHE; specify argument intents
  • key_info.f - get rid of variables NPHI and NTHE, specify argument intents and completely overhaul routine to improve clarity and readability
  • calc_mod_tps.f - rename (NPH, NTH) to (NPHI, NTHE) and declare as local variables rather than subroutine arguments.
  • get_nview_mod.f - declare NTHE as local variable rather than subroutine argument.
  • pmap_fft.f - declare array A to be complex of size NROT rather than real with size 2*NROT and make corresponding code modifications.
  • fft_2d.f - specify argument intents and cleanup comments.
  • fft_2d_back.f - specify argument intents, cleanup comments, and use Fortran 90 array syntax.
  • global_cc.f - declare array PPFT to be complex rather than real; delete flip, index, mode, and sq1 from argument list in call to get_tpo_g; move hand flip calcualtions from here into get_tpo_g.
  • get_tpo_g.f - specify intent for all subroutine arguments; delete flip, index, mode, and sq1 from argument list; move hand_flip calculations here from global_cc; declare pfft to be complex and make necessary code modifications
  • ccf_fft - Complete overhaul for improved clarity, use of complex arrays, cleanup of logic, commenting. In particular, switching from real to complex arrays vastly simplifies index calculations.
  • fill_params.f - specify argument intents, extensive overhaul of logic for improved clarity
  • write_params.f - overhaul and cleanup for improved clarity; specify argument intents.

Summary of changes that result in slight numerical differences

• Faster approximate algorithm for determining ω in program PPFT. Enabled by default and controlled by ‘ppft quick_omega’ parameter.
• Faster approximate algorithm in POR for local search of orientation space. Enabled by default and controlled by ‘por quick_search’ parameter.
• Apodized map produced by P3DR. Width of border region controlled by ‘p3dr apo_border’ variable, default value equal to 12 pixels.
• Automatic refinement of CTF parameters now turned off by default. Controlled by ‘auto refine_ctf’ parameter.
• Bug fix in auto3dem so that inner and outer radii of capsid estimated correctly when binning used in PPFT

Detailed description of changes

• Major performance enhancements made to PPFT routine (get_phiomega) responsible for determining the orientation angle ω and the sign of φ. For smaller problems, this routine accounts for a very tiny fraction of total run time, but owing to the scaling behavior of the algorithm (NROT3, where NROT depends on map size and binning factor), get_phiomega can dominate the run time for larger problems.

The initial search for ω is now done over a coarser grid of values, with the step size dependent on NROT. This is followed by a local search using a finer grid in the vicinity of the top scoring values of ω obtained during the initial search. After several iterations of global search mode, only a small fraction of the particle images have orientations that differ from those obtained using the original algorithm and the actual differences in the orientations are minimal.

The 'quick omega' feature is enabled by default in auto3dem. Adding the following line to auto3dem input files disables this option

ppft quick_omega  0 # auto3dem input file

This new capability required the creation of a new subroutine get_phiomega_quick and modifications to the following files in directory PPFT: global.f, global_cc.f, get_tpo_g.f, pftsearch.f, ppft_info.f, and key_info.f. Also required changes to include/infohead.inc, init_params.pm, and make_program_input.pm

• Capabilities added to P3DR to produce an apodized map, with a Gaussian falloff applied to the density in the region boxrad-border ≤ r boxrad. Border has units of pixels and must be an integer. Using a border width of zero recovers original P3DR behavior. Default value in auto3dem set to 12, but can be overridden by adding the following line to the auto3dem input files

p3dr apo_border n # auto3dem input file

This new capability required changes to the following files: P3DR/P3dr.F, P3DR/density_clear.f, Commpk/info.f, Compar/bcast_parameters.f, info.inc, init_params.pm, and make_program_input.pm

• Provide 'quick search' capabilities in POR so that a restricted local search of orientation space (one Δq step along each direction for each orientation angle) is performed first and only those particle images that find a better orientation in this restricted region are subjected to a more extensive local search. Using the 'quick search' feature typically results in a very small fraction of the particle images settling into orientations that are different from those obtained using the more extensive local search, but has been shown to decrease run times by up to a factor of five.

The 'quick search' feature is enabled by default in auto3dem. Adding the following line to auto3dem input files disables this option

po2r quick_search 0 # auto3dem input file

This new capability required changes to the following files: POR/Por.f, POR/cmpt_ort.f, Commpk/info.f, Compar/bcast_parameters.f, info.inc, init_params.pm, and make_program_input.pm

• Refinement of CTF parameters using program PCTFR is now disabled by default. To override, add the following line to auto3dem input file

auto refine_ctf 1

• setup_rmc.pl prints comments to the auto3dem input file describing how to set parameters to perform faster reconstructions and reach higher resolutions. The following commented out parameters are now listed

ppft bin_factor
ppft verbose
ppft annulus_low
ppft_annulus_high
pcut in_rad
pcut out_rad
auto freeze_annulus
auto bin_reduce

• Bug fix made to Perl module get_ann_lo_hi.pm so that inner and outer radii of capsid are estimated correctly when binning is used in program PPFT.

• POR modified to keep statistics on both the level of individual micrographs and the entire run for the number of particle images that fall into each of the following categories

Stable orientation / stable origin
New orientation / stable origin
Flipped hand / stable origin
Stable orientation / new origin
New orientation / new origin
Flipped hand/ new origin

The 'stable' designation means that the orientation or origin for a particle is unchanged after running POR. 'Flipped hand' means that the orientation is the same except for a change in the handedness of the particle. This information is summarized in the POR output file.

Extra fields have been added to the new particle parameter files generated by POR so that the status of the orientation (Stable, New, Flipped) and origin (Stable, New) is listed for each particle.

• Add new Perl module to auto3dem to perform tests that determine whether or not particle parameter files are properly formed. Affects files auto3dem.pl and sanity_checks.pm

• Allow for specification of P3DR parameters magfactor (default value 1.0) and map_dim (default value 0) in auto3dem input file. Affects files init_params.pm and make_program_input.pm

• Modify PCTFR/cmpt_ctf.f to handle arbitrary formats in the particle records of the particle orientation files. Records are now read in as strings, rather than parsed into individual fields, and written out to the new particle file following the updated CTF parameters. This allows us to include additional fields that may be written by programs POR or PPFT without having to make modifications elsewhere in the code.

• Command line syntax for auto3dem and setup_rmc more flexible. Key value pairs no longer need to be separated by equal signs. For example, both of the following are now valid

%auto3dem -ncpu=16 -input=input_file
%auto3dem -ncpu 16 -input input_file

• Minor performance enhancements made to (q, |φ|) search routine get_thephi in program PPFT.

• Make following modifications to PPFT input routine ppft_info:
  • Equal sign in key-value pairs is now optional
  • Strings containing ‘/’ properly handled
  • lg_out initialized to zero
  • Start of data file list can be specified using both ‘end_of_keys’ (first three characters significant) or ‘inputparameterfiles’ (first nine characters significant).

• Allow blank lines and comments (using ‘#’) in the data file section of input for programs PPFT, POR, P3DR, and PCTFR. Required changes to PPFT/pfile_info_pft.f and Commpk/pfile_info.f.

• The remaining modifications listed below for v2.02 all involve internal code changes that do not affect results, use of codes, or specification of input parameters.

• Cleanup timing functionality throughout code
  • Add new function elapsed_time.f to library Commpk that calculates the time elapsed between pairs of system_clock readings; get rid of old timer function timer1.c, and update Makefile.
  • Modify mem_time.inc so that all timing variables reside in a common block and get rid of test_memory, test_time, and wtime declarations.
  • Replace calls to mpi_wtime with Fortran 90 system_clock intrinsic in the following files: Ctffit/ctffit.f, P3DR/P3dr.F, P3DR/cmpt_intrps.F, PCTFR/Pctfr.f, PCTFR/cmpt_ctf.f, PCUT/Pcut.f, POR/Por.f, POR/cmpt_ort.f, PPFT/global.f, PPFT/global_cc.f, PSF/Psf.f
  • Get rid of old commented out code in following files: P3DR/matrixentries_slab.F, PCTFR/Pctfr.f, POR/Por.f, POR/cmpt_ort.f.
  • Comment out or remove unused variables and labels in following files: PCTFR/cmpt_ctf.f, PPFT/list_ccs.f, PPFT/pfile_info_pft.f, PPFT/ppft_info.f, Ctffit/ctffit.f, Ctffit/rotate.f

• Modify programs PSF, PCUT, P3DR, PCTFR, PPFT and POR to allow specification of input file both as command line argument and through redirection of standard input. For example, both of the following are valid ways to execute P3DR

% mpirun –np 8 P3DR p3dr_input > p3dr_output
% mpirun –np 8 P3DR < p3dr_input > p3dr_output

The first syntax is preferred and is used internally by auto3dem when launching the parallel jobs. The reason for making this change is that many MPI implementations cannot handle redirection from standard input if the number of characters is greater than 4096. The second syntax is retained for continuity purposes only and is no longer recommended.

• Add full path to nodefile in setup_rmc and auto3dem.
• Add capabilities to change search mode bin factor in auto3dem using tests analogous to those used in making decision to switch from search mode to refine mode. Controlled through ‘auto bin_reduce’ parameter. Bin factor is now listed in summary file.
• Fix logic error in auto3dem that resulted in mode (search à refine) being updated after the restart file was written. This bug did not affect results of image reconstruction, but could result in an extra iteration of search mode being carried out if the calculations died during the first iteration following the automated transition from search to refine mode.
• Added new script config_test.pl to main code directory to perform simple tests on the computing environment. Display Perl version is displayed and determine whether or not all required and optional Perl modules are present. Confirm that mpirun, mpicc, mpif90, and scalar C and Fortran compilers called by mpif90 and mpicc, respectively, are in path.

The remaining modifications listed below for v2.01 all involve internal code changes that do not affect results, use of codes, or specification of input parameters.

• Eliminated temporary array PRJSL in file PPFT/global.f
• Cleanup and reorganization of PIFlib/libPIF.c

Important - Version 2.0 contains a new command line interface for running setup_rmc and auto3dem. The old usage syntax is no longer valid. Any scripts that call setup_rmc or auto3dem must be modified to use the new syntax.

• Completely overhauled command line interface for both auto3dem and setup_rmc so that input is done using key-value pairs. For example

setup_rmc -ncpu=4 -seed=123 -list=listfile
setup_rmc –usedefaults

auto3dem -ncpu=12 -input=input_file -nodefile=mynodelist

setup_rmc can be run purely with default values using the –usedefaults flag. For auto3dem, values must be provided for the number of CPUs and the name of the input file. The nodefile no longer needs to be specified for either setup_rmc or auto3dem. This is to be contrasted with previous versions where the word 'none' had to be explicitly used for those cases where the parallel computing environment did not require a node list. For batch systems running the PBS scheduler, the PBS node file is automatically obtained from the $PBS_NODEFILE environment variable via Perl’s %ENV hash.

Both programs list usage information if executed either without any arguments or with the -help flag. All flags are case insensitive and whitespace before or after the '=' in the key value pairs is tolerated. In addition, setup_rmc prompts the user to continue run using all default values after printing usage information.

• Major performance improvements in P3DR for icosahedral symmetry. Twofold and threefold symmetry operations are applied to 3D DFT of model, thereby reducing the number of interpolations required for each image from 60 to 5. Speedup is problem and system dependent since it depends on the number of particle images and the relative time required for different operations (FFT, interpolation, etc.), but runs times have been measured to be reduced by 8x for test cases involving 1000 particle images.

• Added capability to handle both relative and absolute path names to boxed image files (as specified in particle parameter files) to setup_rmc.

• CTF corrections can be turned off globally using -noctf option on the setup_rmc command line and 'auto noctf 0' in the auto3dem input file. This has the same effect as manually setting the CTF mode for programs P3DR, PO2R, PCTFR, and PPFT to zero. Using the -noctf option in setup_rmc automatically propagates setting of the noctf flag in the generated auto3dem files.

• The auto3dem input file now recognizes the keys ctf_mode and ctfmode for all programs that perform CTF corrections. This change was made to remedy the confusing situation where PPFT used ctf_mode while P3DR, PO2R, and PCTFR used ctfmode.

• Intermediate files generated during random model method calculations are all moved to a new directory named RMC_temp. The best starting model is renamed rmc.pif and moved into the directory containing the particle parameter files. Files and directories corresponding to random orientations that are not used are automatically deleted using a new utility script, remove_useless.pl, which is called from RMC_run. A short script named RMC_cleanup is generated by setup_rmc for the purpose of recursively removing contents of RMC_temp.

• setup_rmc automatically generates a basic auto3dem parameter file to continue reconstruction after random model method calculations have been completed.

• Sensible defaults now used for obtaining list of particle parameters files considered by setup_rmc. If -list key is not used, setup_rmc first looks for a file named ‘list’. If this file does not exist or is not readable, then the data directory is queried for files of the form *000, *001, etc.

When a list of parameter files is read from a file, now have added flexibility in the specification of the files. Wildcards, comments, blank lines, and whitespace are now allowed. Can also use a single quoted expression in place of the file. For example, -list=’*001’ will be expanded within setup_rmc to a list of all files ending with 001 in the specified directory.

• Modified auto3dem.pl so that GD::Graph module is loaded at run time using 'require' rather than 'use'. This makes it possible to trap exceptions and bypass the FSC graph generation if the module cannot be found. The advantage of this approach is that the auto3dem.pl source does not have to be manually edited to handle Perl installations that are missing GD::Graph.

• PCTFR is now called by auto3dem only if ‘auto mode’ equals refine and both ‘pctfr ctfmode’ and ‘auto refine_ctf’ are true (non-zero).

• Lower memory algorithm and single precision interpolation option for P3DR have been removed. This both simplifies the source code and makes it easier to implement planned performance enhancements. The default is to use the more memory intensive (and less communications intensive) algorithm and double precision interpolation options, but these could previously be overridden using the –DALG3 and –DSPREC compiler flags.

An older version of P3DRhas been retained in the P3DR_old directory, but does not contain the performance enhancements described above.

• Semicolons now allowed, in addition to commas and whitespace, as delimiters in list of email addresses in auto3dem input file. For example,

auto recipient address1 address2, address3; address4

• MPI communications in PPFT/global.f modified to avoid exceeding MPI buffer limitations. Allows PPFT to be run using a finer spacing between projections of the model and/or larger models.

• MPI communications in P3DR/exch_intp.F modified to avoid exceeding MPI buffer limitations. Allows P3DR to be run for larger image sizes.

• PSF/Psf.f modified to give warning message rather than terminating execution if the pixel sizes in the even and odd map headers are different.

• PCTFR/Pctfr.f modified to use pixel size as specified in the particle parameter files rather than the map files. Warning messages are provided if the pixel size of the map is either negative or larger than the particle parameter file pixel size.

The remaining modifications listed below for v2.0 all involve internal code changes that do not affect results, use of codes, or specification of input parameters.

• Variable r4_max set to huge(r4_max) in PCTFR/cmpt_ctf.f and POR/cmpt_ort.f
• General cleanup and improved commenting of Compar/bcast_parameters.f and Compar/read_files.f
• Remove nodefile from validate_commandline_args() argument list in validate.pm and from call to function in auto3dem.pl. Allow nodefile to be undefined in auto3dem, writer_header, and run_mpi_prog.

• Default values are now set for the 'auto outfile' and 'auto rundir' parameters. Unless otherwise specified, outfile is set to the name of the working directory where auto3dem is launched and rundir is set to be the directory dat. This involved changes to the file init_params.pm.

• Modified Fortran input routines pfile_info and pfile_info_pft to handle more flexibility in the specification of the image file names in the particle parameter files. The image name is first tested exactly as entered in the parameter file. If the program is unable to open it, the path name to the file is stripped off and an attempt is made to open the file in the current working directory. Similar changes made in the Perl module get_info.pm.

• Rename the auto3dem parameters fsc_locut and fsc_hicut to fsc_lothresh and fsc_hithresh, respectively. Scripts will still accept the old parameter names, but will be converted internally to new names. These changes were made so that that the names of the variables would more accurately reflect their usage as thresholds in the FSC data. Similarly named variables in the Perl code were renamed to use the _lothresh and _hithresh suffixes. Also reordered return argument list from PSF parsing routine. These changes involved modifications to the files init_params.pm, auto3dem.pl, psf_parse.pm, and updatre_res.pm.

• Minor bug fixes to the files Commpk/ctf_para.f, POR/Por.f, and P3DR/P3dr.F to properly handle the case where no CTF correction is applied. PCTFR/Pctfr.f modified so that it exits early if ctfmode is set so that no correction is applied. Commpk/info.f fixed to properly handle num_den_pix set equal to zero in the input files.

The remaining modifications listed below for v1.12 all involve internal code changes that do not affect results, use of codes, or specification of input parameters.

• Intent of interpolants argument in P3DR/cmpt_interps.F changed to INOUT.

• In the following routines, replace the variable stdoutput with * in write statements: Commpk/intlz_arrays.f, Commpk/symmcode.f, P3DR/P3dr.F, P3DR/exchange_2_slab.f, P3DR/fftsynth_1_m_slab.f, P3DR/fftsynth_1a.f, PCTFR/Pctfr.f, PCUT/Pcut.f, and POR/Por.f

• Completely overhaul the include file include/allprog.inc. Get rid of stdinput, stdoutput, max_input, pi, twopi, deg_to_rad, and rad_to_deg; use Fortran90 parameter statements to set constants; improve comments; move filename_len from allprog.inc to info.inc. Get rid of variable skipone since it is used only in readorient.f and declared with SAVE attribute to retain value between calls.

• Cleanup and overhaul include file include/info.inc. The variable filename_len is now declared here so that the use of info.inc does not rely on allprogs.inc.

• Get rid of skipone variable in Commpk/info.f

• Replace filename_len with hardcoded value 257 in files Compar/read_map.f and Ctffit/ctffit.f

• Get rid of oddeven parameter in POR/cmpt_ort.f and PCTFR/cmpt_ctf.f.; in calls to readorient, replace with zero.
• In routine Commpk/intlz_params.f get rid of initialization of pi, twopi, rad_to_deg, and deg_to_rad

• Overhaul logic in PPFT routine involved with the VAROPT feature. This is an experimental feature that is currently not recommended, but code modifications may have fixed a long time bug. Changes involved files global_cc.f, pftcc_fill_g.f, get_tpo_g.f, avg_pftimg.f, and get_thephi.f. Also declare variable pi locally in pftsearch.f

• Email notification feature modified so that messages can be sent to multiple email addresses and that an arbitrary number of files can be attached. Individual addresses may be separated by whitespace and/or commas and each individual address is tested to be well formed. Note that additional fields in auto3dem parameters file records are still ignored except for the 'auto recipient' record. Modifications involved files auto3dem.pl, sendmail.pm, and init_params.pm

• Capabilities added to generate and email graphs of FSC curves using GD library. Since not all sites have the required GD and GD::Graph Perl modules, the code in auto3dem.pl marked with the #GDSFC to construct the graphs must be manually uncommented.

• Input files info.f and pfile_info.f in library Commpk completely overhauled. Can now handle P3DR and PO2R input files that contain blank lines, tabs and leading whitespace.

• OPEN statements in key_info.f and write_params.f modified so that PPFT no longer dies if it tries to overwrite existing files.

• Added subroutine indexx (borrowed from Numerical Recipes in Fortran90, with very minor modifications) to the Commpk library. Will be used in future releases to sort records in particle parameter files by their ID.

• PCTFR/cmpt_ctf.f and POR/cmpt_ort.f remove all whitespace from particle image file names before writing particle parameter files.

• POR/cmpt_ort.f now writes zeroes for the last two fields in the particle parameter files. By making this change, the files generated by PPFT and POR will have the same format.

• Commpk/readorient.f modified so that it can deal more reliably with particle parameter files that are missing the last two scores. These types of files are generated by OOR and older versions of PO2R.

• File Commpk/read_1_pif.f cleaned up and commented. Include statements removed and argument list expanded. Corresponding changes made to argument list in Por.f, P3dr.f, and Pctfr.f.

Minor changes made to auto3dem so that:

• Temporary files stdin_temp and message.txt are deleted
• Log and summary files are properly handled in the event that a new run is launched but the log and/or summary files already exist
• FSC data and graphs labeled using same prefix as map, summary, log, and restart files.

• setup_rmc no longer specifies the number of models on the command line. New syntax is

setup_rmc dir list ncpu nodefile [boxrad]

• Streamlined random model calculations – auto3dem.pl, init_params.pm, write_params.pm, setup_rmc.pl, and findbest.pl were modified to so that calculations on a particular random model are terminated once the resolution improves to the point where the FSC curve never drops below 0.5. Also, additional random models are not calculated once one of the random models had successfully converged. In the best cases, this results in a 20x reduction in the run time relative to calculating ten full iterations for ten different random models.

• Maps generated by auto3dem have a better naming convention, using the 'auto outfile' parameter to specify the file prefix and '.pif' to specify the suffix. For example, the 5th map generated where outfile equals reovirus will now be named reovirus_iter_5.pif rather than map_iter_5. Script findbest.pl modified to handle new naming conventions.

• Improved naming of variables in several of the Perl scripts and modules.

• Always output memory usage and timings in P3DR, PCTFR, PCUT, and POR. These had been under control of the hard-coded logical variables test_time and test_memory.

• Get rid of calls to bcast_parameters in Psf.f and Pcut.f. This makes the code easier to follow since bcast_parameters broadcast parameters that were listed in a common block and allows for more sensible naming of variables.

• Get rid of 'include ../info.inc' in PCUT/cut_map.f and expand argument list. Get rid of 'include ../mem_time.inc' in Pcut.f and declare necessary variables rather than relying on common blocks.

• Completely overhaul logic in PSF to make the sequence of operations easier to follow.

• Email notification – a new module sendmail.pm was added to AUTO3DEM that provides capabilities for sending email notifications with optional attachments. This feature requires that the standard Linux/UNIX mutt email tool be installed on your system. The sendmail.pm module tests for mutt using the UNIX ‘which’ command and does a cursory check on the email address to make sure that it meets the minimal requirements for being well formed. auto3dem.pl has been modified to send email notifications, with a text output of the FSC calculations, at the end of each iteration. To enable this feature, add the following line to the AUTO3DEM parameter file, where email is a valid email address.

auto recipient email

• Random model computations have been simplified. Modifications have been made to both setup_rmc.pl and findbest.pl so that best map is identified and copied into the directory provided as argument to setup_rmc. Argument list for setup_rmc has been expanded so that the commands file does not need to be hand edited and so that all temporary files and directories associated with random model computations have the ‘RMC’ prefix. This last change makes the cleanup of intermediate files more straightforward. The updated syntax for running setup_rmc is shown below. Note the new ncpu and nodefile arguments.

setup_rmc dir list nmodels ncpu nodefile [boxrad]

• CTF refinement – the program PCTFR (Parallel CTF Refinement) has been integrated into AUTO3DEM. This involved changes to auto3dem.pl, init_program.pm, make_program.pm, and write_param.pm. The subroutine cmpt_ctf called by PCTFR has been overhauled and modified so that it can handle both stigmatic and astigmatic images. PCTFR does not currently handle the transition from a stigmatic to astigmatic images unless a reasonably accurate estimate of the orientation of the astigmatism is provided. The main program Pctfr.f has been cleaned up and modernized.

CTF refinement is controlled by the new parameter ‘auto refine_ctf’ and is enabled by default. Note that CTF refinement is only done when running in refine mode since at lower resolutions the spatial frequencies may not span enough nodes in the CTF function to make a reasonable estimate of the defocus.

• Multiplication of model projections by CTF in PO2R – In previous versions of PO2R, the projections of the model were divided by the CTF function before comparing them to the images. Modifications were made to files Por.f and cmpt_ort.f so that projections of the model are multiplied by CTF. Either approach is valid for comparing the model to the images, but multiplication by the CTF avoids the approximations that are required to properly handle the inverse CTF in the vicinity of the nodes.

• Tests added to AUTO3DEM (module update_res.pm) to ensure that the maximum resolution used in P3DR is not less than twice the pixel size (Nyquist limit).

• Default values now set for PCUT in_rad and out_rad; PPFT annulus_low and annulus_high. It should be noted though that it is still to the user’s advantage to manually enter values for these parameters once estimates become available from lower resolution maps.

• PPFT modified so that existing output files are overwritten rather than having the program terminate. In addition, PPFT can now handle leading whitespace before the names of parameter files and trailing empty lines in the input file.

• General code improvements – removed unnecessary include statements in fft_2dfft; generalized maptempfac to handle arbitrarily shaped arrays; modified readorient to return score and correlation coefficient data from files generated by PO2R and PPFT, respectively; modified cmpt_intps, cmpt_ctf, and cmpt_ort to be compatible with new readorient routine;

• Temperature factor – ctf_para now applies temperature factor in the same manner (positive sign in exponent) for all values of filter. Simplifies logic of calling ctf_para with regards to sign convention. All code (except possibly PPFT) should handle temperature factor calculations correctly.

• Random model pixel binning – changed threshold for using binned data (bin_factor=2) in random model calculation from pixels of size 5Å to 4Å.

• Hollow map bug fix – Repaired bug that affects hollow map calculations when running in search mode. Since hollowed maps are an advanced feature and are normally only used in refine mode, this bug probably had no impact on any reconstructions performed to date.

• PO2R res_min bug fix – Repaired bug so that PO2R minimum resolution is properly calculated. Bug would only have been encountered if user was trying to manually override default value for res_min based on calculation involving pixel size.

• PO2R handedness test bug fix – Repaired bug that had been introduced in version 1.08 regarding test for proper hand.

Modifications dealt primarily with the elimination of redundant functionality. In particular, the use of equivalent_view to generate symmetry-related orientations and crowther_to_matrix to determine rotation matrices corresponding to the orientations defined by (theta, phi, omega). These changes lead to very small numerical differences. A number of non-algorithmic modifications were also implemented that improve code readability.

• Commpk – split the file to_asym_unit.f into separate files each containing a single subroutine or function; got rid of common block and used SAVE attribute to retain the constants in function good and subroutine genrot that were previously calculated in geometry_init. Removed subroutines setrotmat and getorient since these are no longer needed. Moved the routines density_clear and maptempfac from library Compar to Commpk since they do not contain any parallel code.

• Compar – Moved routines density_clear and maptempfac to library Commpk.

• P3DR – Overhauled calculation of symmetry-related orientations for icosahedral symmetry in subroutine cmpt_intrps; replaced a fairly large block of confusing code with calls to routine equivalent_view. Replaced call to setrotmat with crowther_to_matrix.. Updated calling sequence to eight_symmetry and dihedral symmetry to take (theta, phi, omega) as arguments rather than a 3-element vector. Got rid of subroutine ico_vector since it is no longer needed for determination of equivalent orientations.

• POR – Replaced calls to setrotmat with crowther_to_matrix in subroutine cmpt_ort.

• PCTFR - Replaced calls to setrotmat with crowther_to_matrix in subroutine cmpt_ctf.

• PPFT – get rid of files wavel.f, ctf_scale.f ,and piraddeg.f; move ctf_firstpeak_pft.f to library Commpk, and replace calls to ctf_scale with calls to ctf_para in subroutines global_cc and calc_pfts_g. Make minor, non-algorithmic (general cleanup, redefining real arrays as complex arrays, specification of argument intents, etc.) changes to multiple files.

• POR – change one-dimensional arrays to multi-dimensional arrays in Por.f and cmpt_ort.f. This does not change any results, but does make code easier to follow. Get rid of file imgcompcet.f and replace calls to imgcompcet with calls to imgcomport.

• P3DR - change one-dimensional arrays to multi-dimensional arrays in P3dr.F; minor modifications to matrixentries_slab.F.

• PCUT - change one-dimensional arrays to multi-dimensional arrays in Pcut.f

• PCTFR - change one-dimensional arrays to multi-dimensional arrays in cmpt_ctf.f; minor modifications to Pctfr.f

• PSF - change one-dimensional arrays to multi-dimensional arrays in Psf.f

• Compar – general cleanup of exch_3d_1.f

• Commpk – split ctf_para.f file into multiple files ctf_es.f, ctf_et.f, ctf_firstpeak.f, ctf_func.f, ctf_para.f, and ctf_temp.f. Overhaul and cleanup all routines, with an emphasis on making them usable by POR, P3DR, and PPFT. Move PPFT/ctf_firstpeak_pft.f into Commpk. Perform general cleanup of disef_para.f, focus_astig.f, imgcomport.f, intlz_arrays.f, info.f, rpifimag.f, symmcode.f

• Remove files ico_EM4IMR.f and ico_EM4IMR_vector and replace with single routine in file ico_vector.f. New version may lead to very slightly different results since the hard-coded values 0.809017, 0.500000, and 0.309017 have been replaced with the more accurate representations cos(36°), cos(60°), and cos(72°), respectively.
• Modified the code for filter=1 in ctf_para so that CTF is zeroed in when the condition |CTF|<0.1 is met rather than |CTF| < sqrt(ctf_ff2). The old version could possible lead to the CTF being set to zero for a very large range of CTF values.
• Removed the include file vax_minmax.inc since it is no longer needed by any routines.

Minor, non-algorithmic changes made to many files in Commpk, Compar, P3DR, POR, PPFT, PSF, include, and Ctffit directories. Primarily removal of unused variables, specification of argument intents, improved comments, etc.

• Improved capabilities added for exiting auto3dem if one of the image processing programs aborts. Depending on operating system and batch queuing system, the return value from launching an MPI job with system command may not be accessible to the auto3dem script. To ensure that the abort is detected, the output from the MPI program is parsed and auto3dem exits if the string MPI_Abort (not case sensitive) is found. Involved modifications to error_stop.f, run_mpi_prog.pm, and auto3dem.pl.
• Minor performance improvements made to 3D interpolation routine interpl_3d.
• General cleanup of cmpt_intrps.F, fftsynth_1_m_slab.f, exchange_2_slab.f, P3dr.F, Pcut.f, ctf_para.f, Psf.f, Por.f, Pctfr.f, com_01.inc, intlz_arrays.f, and intlz_params.f. Specification of argument intents, removal of unused variables, etc.
• General cleanup of PSF/comp_sfactor.f. Move calculations contained within subroutine cplex2ap into comp_sfactor and get rid of unnecessary operations.
• New control parameters added:
  • auto switch_mode (see note below)
  • auto term_refine (functionality not yet active)
  • auto term_search (functionality not yet active)
• Auto3dem now monitors resolution and can automatically switch from search to refine mode if the following conditions are met:
  • switch_mode flag is true
  • Resolution has not improved by at least 0.25Å over previous iteration
  • Mode is currently search
• Iteration information is now tracked so that numbering from one run to the next is maintained. Restart and continuation files specify the last iteration that had been completed. The restart files also keep track of the number of iterations required to finish the original calculation.
• For restart or continuation, new results are appended to the log and summary files.
• Restart files are written at two points in each iteration: after the completion of origin and orientation refinement and after construction of map. The restart files properly set the auto have_map flag so that map is constructed from particle images if program needs to be restarted after first checkpoint.
• Output files are named consistently using the name of the auto outfile parameter:
  • outfile_log – detailed output log
  • outfile_summary – summary information
  • outfile_restart_na – restart after first checkpoint, iteration n
  • outfile_restart_nb – restart after second checkpoint, iteration n
  • outfile_continue – continuation file for successfully completed runs
• Summary file now lists the number of particles that had been selected to construct the final map and the number of CPUs (MPI processes) used.
• Format of restart and continuation files has been improved to logically group parameters.
• Default value (0.1 pixels) specified for po2r dcenter, step size for origin refinement.
• auto final_map no longer used and has been labeled as deprecated.

• Auto3dem version now printed in log and summary files
• Symbolic links to executable Perl scripts in BIN directory created. If path information is setup correctly, can use name of script with or without .pl extension.
  • auto3dem points to auto3dem.pl
  • setup_rmc points to setup_rmc.pl
  • handflip points to handflip.pl
  • findbest points to findbest.pl
• auto3dem and setup_rmc executed without any arguments provides usage information and version
• Improved format for auto3dem summary data
• Fixed bug in auto3dem.pl where PPFT output files were moved before they were no longer needed.
• Minor changes to output from update_res.pm
• Default value for lower resolution limit of FSC curve (psf res_min) set to 60Å
• Default value for PO2R angular step size (po2r dangle) set to one-half of the angular step size for PPFT.
• Default value for PO2R res_min set to 2/5 * boxrad * pixelsize
• Default value for PPFT resolution_low set to 2/5 * boxrad * pixelsize
• Test provided on number of iterations in input file. Must be defined and be at least zero.
• Test added to make sure that at least one ‘data’ line is specified and that if wildcards are used they expand to include at least one file.

Changes were implemented that make it much easier to run AUTO3DEM. Rather than creating symbolic links to Perl code and binaries, path information is simply added to the computing environment. This required some changes to the Perl code so that binaries and custom modules would always be found.

• Custom Perl modules reside in the same directory as the executable Perl scripts. Added the following line to all .pl and .pm files so that custom modules would always be found, regardless of where the scripts are executed

use lib do { __FILE__ =~ m|^(.*?)[^/]*$|; "$1"; };

• Modified setup_rmc.pl so that executable Perl scripts in the command file are not pre-appended with './'. For example

./auto3dem.pl   becomes   auto3dem.pl

• In module init_params.pm, get rid of full path information for binaries

• Modified so that all binaries are copied into the BIN directory

auto3dem

• Modified auto3dem.pl and init_params.pm so that map can be constructed without doing resolution estimation. This feature is particularly useful when applying inverse temperature factor to final map.
• Particle selection criteria can now be applied globally across all files or on a per file basis. By default, selection criteria applied globally. Choice controlled by auto global_select.
• Create subdirectory for storing the program input and output files, together with the filtered particle parameter files generated by applying the particle selection criteria.
• Rename the PSF output file corresponding to the best particle selection criterion for a given iteration to FSC_curve_{iteration}.
• Handle the case where the FSC curve never drops below fsc_locut. Particular useful in the early stages of search mode where resolution of map has the most rapid improvements.
• Get rid of dryrun option
• Write summary file containing just the most important information required to monitor the progress of the reconstruction.
• Modify init_params.pm to ignore ‘auto dryrun’ and ‘auto split_ppft’ and provide message that these are deprecated options.

Program POR

• Modified Por.f so that program can deal with map files that contain ridiculous values for the pixel size. In this case, the pixel size from the particle parameter file is used instead.

handflip.pl

• Script for changing the handedness of orientations in particle parameter files.

Program P3DR

• Interpolation routine matrixentries_slab.F improved to yield better performance. Actual speedups will vary by problem and hardware, but runs using test problem show 28% reduction in time spent in 2d interpolations and a 15% reduction in overall run time.

Program P3DR

• Modified P3dr.F and get_1st_ortid.f to properly handle empty particle parameter files.

Program PPFT

• Added option for verbose=-1, which results in the correlation coefficient cc_cmp not being calculated. This option should only be used for construction of starting model using random model method since cc_cmp is generally used to select particle images.

setup_rmc

• Script setup_rmc.pl modified so that default values for PPFT are overridden.

delta_theta = 1° (default 0.5°)
verbose = -1 (default 2)

auto3dem

• parameter file now accepts po2r for specifying PO2R input. por still works to allow back compatibility with v1.0, but all internal perl variables renamed to use po2r (e.g. %por_ref renamed internally to %po2r_ref)