Generating a starting model
Obtaining a reconstructed density map from a set of particles is an iterative process that requires an initial reference volume, typically a low-resolution reconstruction from the same or a similar sample. When a starting model is not available, one can easily be constructed for icosahedral viruses using the program setup_rmc, which relies on the Random Model Computation (RMC) method.
Assuming that you have already generated a dat subdirectory containing the initial particle parameter files (ending with .dat_000
), the simplest procedure is to run three commands
setup_rmc ./RMC_run ./RMC_cleanup
setup_rmc will interrogate the particle parameter files in the parameter directory (default=dat) and setup the directories and scripts required to generate, by default, 10 random models, using 150 images from at least the 3 furthest-from-focus micrographs. It will generate two scripts to run in sequence: RMC_run to launch the random model calculations, and RMC_cleanup to remove all intermediate files after completion. setup_rmc will also construct a minimal auto3dem input file (with suffix _master
), that can be used to launch the full reconstruction after the starting model has been obtained.
The number of random models to construct, the number of images to use, and other parameters can be controlled through command line arguments to setup_rmc
. Running setup_rmc without any arguments provides a complete list of options.
For example, to create 5 random models using 200 images taken from a minimum of 4 micrographs
setup_rmc -nmodels 5 -nimages 200 -nmgmin 4
By default a novel algorithm is used to select the best starting model. In this case all the models are run for the number of iterations specified (default = 10), without calculating the resolution after each iteration. A score based on amplitude statistics is calculated from the radial profile of each model, and the volume with the highest score is selected as starting model. If the resolution-based approach is chosen (-trad
flag), at each iteration the resolution is estimated for each model: if the FSC never drops below 0.5 for all spatial frequencies less than 1/25 Å, the calculations are terminated and the current model is designated as the starting model. Otherwise, all calculations are allowed to run to completion, and the best model, according to the FSC resolution estimate, is identified. In either case, the starting model is named rmc.pif and copied to the directory containing the particle parameter files (default = dat). At the end of the calculations, all intermediate files and directories are also moved into a new directory named RMC_temp.
A sample session is shown below:
% ls dat pif % setup_rmc [output not shown - enter yes at prompt] % ls dat RMC2_master RMC6_master RMC_cleanup pif RMC3_master RMC7_master RMC_logfile_list RMC10_master RMC4_master RMC8_master RMC_run RMC1_master RMC5_master RMC9_master Virus_master % ls dat file1.dat_000 file4.dat_000 file7.dat_000 RMC10 RMC4 RMC7 file2.dat_000 file5.dat_000 file8.dat_000 RMC2 RMC5 RMC8 file3.dat_000 file6.dat_000 RMC1 RMC3 RMC6 RMC9 % ./RMC_run % ls dat pif RMC_cleanup RMC_temp Virus_master % ls dat file1.dat_000 file3.dat_000 file5.dat_000 file7.dat_000 rmc.pif file2.dat_000 file4.dat_000 file6.dat_000 file8.dat_000 % ls RMC_temp RMC1 RMC2_RESTARTS RMC5_log RMC8 RMC10 RMC2_summary RMC5_master RMC8_continue RMC10_continue RMC3 RMC5_RESTARTS RMC8_log RMC10_log RMC3_continue RMC5_summary RMC8_master RMC10_master RMC3_log RMC6 RMC8_RESTARTS RMC10_RESTARTS RMC3_master RMC6_continue RMC8_summary RMC10_summary RMC3_RESTARTS RMC6_log RMC9 RMC1_continue RMC3_summary RMC6_master RMC9_continue RMC1_log RMC4 RMC6_RESTARTS RMC9_log RMC1_master RMC4_continue RMC6_summary RMC9_master RMC1_RESTARTS RMC4_log RMC7 RMC9_RESTARTS RMC1_summary RMC4_master RMC7_continue RMC9_summary RMC2 RMC4_RESTARTS RMC7_log RMC_bestmap_summary RMC2_continue RMC4_summary RMC7_master RMC_logfile_list RMC2_log RMC5 RMC7_RESTARTS RMC_run RMC2_master RMC5_continue RMC7_summary % ./RMC_cleanup %ls dat pif Virus_master
If you generate an initial model using the Random Model Computation method, please cite:
Yan X., K. A. Dryden, J. Tang, and T. S. Baker (2007) Ab initio random model method facilitates 3D reconstruction of icosahedral particles. J. Struct. Bio. 157:211-225. (pdf)