DomainFit

Script to autofit domains in Chimerax

Goal: Provided a database of proteins and an isolated electron density, and find the protein that best fits the electron density geometrically

The program works for Linux and MacOS. For MacOS, a bit more modification is needed.

Update: 2024/06/15 It seems like parsing using PAE is a lot better. From AFDB, you can download the json in the same folder as the PDB. With ColabFold prediction, use the copy_colabfold_predictions.py. With AF3, it is a bit tricky. You can use the rank_AF3_models.py and copy_AF3_predictions.py but it is not as straight forward to use.

Software and Package Requirements

1. Phenix v1.21 or higher (https://phenix-online.org/)

    phenix.process_predicted_model
   

2. Chimerax v1.5 or higher (https://www.cgl.ucsf.edu/chimera/)

    chimerax.fitmap
   

3. Python 3 or higher

    BioPython
    pandas< 2.0, numpy < 2.0
   

4. Rscript 3.6.3 or higher (https://cran.r-project.org/)

    fdrtool
    psych
   

Installation

DomainFit Python Script

git clone https://github.com/builab/DomainFit.git
cd DomainFit
chmod +x install.sh
./install.sh

For MacOSX, you have to modify the "chimerax_path" variable to your ChimeraX's binary file. e.g. /Applications/ChimeraX-1.5.app/Contents/MacOS/ChimeraX in "save_domains_from_info.py" and "fit_domains_in_chimerax.py" and "load_tophits_in_chimerax.py" For Linux, install Biopython and numpy using pip

 pip install numpy
 pip install biopython
 pip install pandas

For MacOS and some Linux, you have to install pandas inside ChimeraX Inside MacOS, open the terminal and type:

/Applications/ChimeraX-1.7.1.app/Contents/bin/python3.11 -m pip install pandas

In Linux, look for the installed ChimeraX folder and open the terminal and type:

/usr/lib/ucsf-chimerax/bin/python3.11  -m pip install pandas

R with fdrtool & psych

 Download R from https://cran.r-project.org/ and install
 Start R in a terminal
 $R
 >>install.packages('fdrtool', repos='http://cran.us.r-project.org')
 >>install.packages('psych', repos='http://cran.us.r-project.org')
 >> quit()
 Save workspace image? [y/n/c]: n

Before usage

source DOMAIN_FIT_DIR/source_env.sh

Workflow

Scripts:

• getAlphaFoldPDBs.py
• process_predicted_models.py
• process_predicted_models_using_pae.py
• save_domains_from_info.py
  • save_domain_single_from_info.py
• fit_domains_in_chimerax.py
  • fit_in_chimerax.py
  • pval_from_solutions.R

Other utility scripts:

• getAlphaFoldPAEs.py
• retrieve_fasta_from_uniprot.py
• copy_colabfold_predictions.py
• plot_domain_histogram.py
• filter_solution_list.py
• clean_up_solutions.py
• write_domain_info.py
• load_tophits_in_chimerax.py
• generate_Rplot.py

NOTE: Try out the example to test the workflow and correct installation.

1. Fetching AlphaFold PDBs

Downloading from AFDB from a list of UniprotID

Input: A list of UniprotIDs (1 per line) in text format (.txt or .csv)

Output: A download directory containing pdbs downloading from alphafold.ebi.ac.uk

Note: The list_proteins.csv must be free of special characters otherwise the program will complain about Unicode utf-8 error. The best way to prepare the file is to copy from a list to a text editor such as gedit.

getAlphaFoldPDBs.py --ilist list_proteins.csv --odir pdb_files

missingAF.log file will tell you which proteins do not have AlphaFold structure available.

If you also want to get the PAE files for domain parsing using PAE graph later

getAlphaFoldPAEs.py --ilist list_proteins.csv --odir pdb_files

2. Parsing AlphaFold predicted models into domains

Generating domain pdbs: Parse automatically using phenix.process_predicted_model, can use default parameters or options.

Input: AlphaFold predicted pdbs

Output: Domain-parsed pdbs (domains.pdb) and domain information files (.domains)

Note: Unfortunately, there are many *_remainder.seq file produced in the main folder. They will be deleted by the script at the end

process_predicted_models.py pdb_files domains nocpu

Alternatively, you can also generate domains using PAE file (https://github.com/tristanic/pae_to_domains). Change the pae_graph_resolution from 1 to 5 for more domains.

process_predicted_models_using_pae.py pdb_files domains nocpu pae_graph_resolution=1

3. Save each domain into single PDB file

Saving domain-separated pdbs using info from the previous step: Parse through domain info and save domains into individual pdbs. Additional screenshots are taken of each domain.

Input: Domain-parsed pdbs, domain_info files folder

Output: Domain-separated pdbs + pngs

save_domains_from_info.py pdb_files domains single_domains minResidueNo maxResidueNo nocpu

4. Fit all domains into ChimeraX

Fitting using ChimeraX: Take each domain and fit it into the density automatically using ChimeraX built in function fitmap. Output pdb saves the fit orientation. Additional screenshots are taken of each domain fitted into the density. A CSV file is generated to document all hits and their respective values (correlation, overlay, overlap, etc.). The default map level is set to 0.034 and a map resolution of 5 Å. Make sure to change this value to the corresponding value for your density.

Input: Domain-separated pdbs + protein density

Output: Solutions with Best-hit pdbs + pngs + csvs

fit_domains_in_chimerax.py inputDir outputDir inputMap mapLevel resolution searchNo noProcessor

Additional Scripts

clean_up_solutions.py

Clean up the solution after having a look at it

Input: Best-hit pdbs + pngs + csvs

Output: Clean up the solution folder keeping only top hits

clean_up_solutions.py solution_dir numberTopHitsRetained

plot_domain_histogram.py

Visualizing domain features: Generates histograms based on the number of residues per domain

Input: <processed> directory

Output: histogram

plot_domain_histogram.py domain_info_dir

load_tophits_in_chimerax.py

Generate a .cxc file to load the top hits for visualization in ChimeraX.

Input: density + solution_dir

Output: .cxc file to open in ChimeraX

load_tophits_in_chimerax.py density solutions_dir number_of_top_hit minsize

filter_solution_list.py

Generate a new .csv file with a minimum size filtering

Input: solution_dir/fit_log_revised.csv file

Output: filtered csv file

filter_solution_list.py solutions_list minsize

visualize_fit_stats.py

Generate a new .eps plot of the fitting statistics from one fitting list with size filtering

Input: solution_dir/fit_log_revised.csv file, minimum size in amino acids (default = 0)

Output: solutions_dir/summaryplot.eps

visualize_fit_stats.py solutions_list minsize

visualize_solutions.py

Generate a visual of top solutions from many searches

Input: solution directories of many densities, a .txt file containing the list of those directories

Output: eps file showing

visualize_solutions.py solutions_csv_list cutoff_rank min_size outputPlot

getAlphaFoldPAEs.py

Fetching AlphaFold predicted alignment error from a list of Uniprot IDs. Not used now but might be useful for other methods of domain parsing.

Input: A list of Uniprot ID (1 per line) in text format (.txt or .csv)

Output: A download directory containing PAEs downloading from alphafold.ebi.ac.uk

getAlphaFoldPAEs.py --ilist list_proteins.csv --odir pdb_files --ignore_existing

generate_Rplot.py

Generate PDF for the R-plot to see the statistical calculation for top hit domains.

Input: solution_dir/fit_log_revised.csv file, number of top hits to plot.

Output: A PDF file for each domain in the solution folder

generate_Rplot.py solution_dir/fit_log_revised.csv 10

Citation

If you use DomainFit for your work, please cite the following preprint:

Gao, J., Tong, M., Lee, C., Gaertig, J., Legal., T., Bui, K.H., (2024) DomainFit: Identification of Protein Domains in cryo-EM maps at Intermediate Resolution using AlphaFold2-predicted Models. Structure. Doi: https://doi.org/10.1016/j.str.2024.04.017