Documentation created by Brahm Yachnin (brahm.yachnin@rutgers.edu), Khare laboratory, and Chris Bailey-Kellogg (cbk@cs.dartmouth.edu). Last edited May 24, 2021.

Purpose

The mhc_energy_tools are companion Python scripts for use with the MHCEpitopeEnergy design-centric guidance scoreterm. In brief, the mhc_score.py script is used to analyze existing protein sequences, PDB files, or lists of peptides to identify immunogenic hotspot sequences. The mhc_gen_db.py script is used to generate a SQL database based from a FASTA or PDB file, with various methods of indicating designable residues and their identities. These databases can then be used to score poses while packing with a mhc_energy term and the MHCEpitopePredictorExternal predictor (see MHCEpitopeEnergy for more details).

Citation information

If you use the Python tools, either in the context of mhc_epitope or independently, please cite the appropriate papers as described here.

The MHCEpitopeEnergy energy method, mhc_epitope scoreterm, mhc energy tools, and benchmark are described in the following publication. If you make use of them, please cite the following paper:

Yachnin BJ, Mulligan VK, Khare SD, and Bailey-Kellogg C. (2021). MHCEpitopeEnergy, a flexible Rosetta-based biotherapeutic deimmunization platform. J Chem Inf Model 61(5):2368-2382. doi: 10.1021/acs.jcim.1c00056. https://pubmed.ncbi.nlm.nih.gov/33900750/

In addition to the Rosetta functionality, mhc_epitope makes use of several de-immunization resources developed outside of the Rosetta community. As a condition for using this code, it is imperative that the resources that you use are appropriately cited in any resulting publication.

ProPred

The ProPred matrices are provided in the Rosetta database (/main/database/scoring/score_functions/mhc_epitope/propred8.txt) and are used by the "default" configuration file, propred8_5.mhc.

The matrices are obtained from http://crdd.osdd.net/raghava/propred/ If you use results based on these predictions, please cite Singh, H. and Raghava, G.P.S. (2001) ProPred: Prediction of DR binding sites. Bioinformatics, 17(12):1236-37. http://www.ncbi.nlm.nih.gov/pubmed/11751237

NetMHCII

NetMHCII can be incorporated into external databases using the mhc-energy-tools. If you use NetMHCII, please cite Improved methods for predicting peptide binding affinity to MHC class II molecules. Jensen KK, Andreatta M, Marcatili P, Buus S, Greenbaum JA, Yan Z, Sette A, Peters B, Nielsen M. Immunology. 2018 Jan 6. doi: 10.1111/imm.12889

IEDB

IEDB is a public database of experimentally-validated immune epitopes. The iedb_data.mhc file references the database file iedb_data.db that contains data derived from the IEDB. If you use either of these files, you must cite the IEDB. In addition, the mhc-energy-tools provides resources to update and build custom database files based on IEDB data.

Please see the following instructions from the IEDB related to using any of these resources:

Users are requested to cite the IEDB when they present information obtained from the IEDB: http://www.iedb.org. The journal reference for the IEDB was updated in 2018 and should be cited as: Vita R, Mahajan S, Overton JA, Dhanda SK, Martini S, Cantrell JR, Wheeler DK, Sette A, Peters B. The Immune Epitope Database (IEDB): 2018 update. Nucleic Acids Res. 2019 Jan 8;47(D1):D339-D343. doi: 10.1093/nar/gky1006. PMID: 30357391

All publications or presentations referring to data generated by use of the IEDB Resource Analysis tools should include citations to the relevant reference(s), found at http://tools.iedb.org/main/references/

mhc_score.py

Input sequences

mhc_score.py can take various types of input for scoring.

  • A sequence can be provided as a positional argument on the command line. For example, mhc_score.py ILVEQACFPSL.
  • --fa mysequence.fas names a file in FASTA-ish format with a SINGLE sequence to be used as input. The > title line is optional.
  • --fsa sequences.fas names a file in FASTA-ish format with MULTIPLE sequences to be used as input. The > title line is mandatory for each sequence.
  • --pdb myprotein.pdb names a PDB file, from which the sequence will be pulled and used as input.
    • --chain X specifies which chain of the PDB file to look at. If not specified, the entire PDB file will be scored.
  • --pep peptides.txt names a file with one sequence per line. Each line will be used as an input sequence.
  • If no input is given, the user can type individual sequences into the terminal. Each sequence will be scored before the next sequence is input.

Additional notes:

  • Only uses one source for sequences, checking in order: command-line, specified input file, stdin
  • Sequences can include '_' characters, treated as noncanonicals, which in the current implementation forces the containing peptides to be non-epitopes
  • The "fasta-ish" header can end in "@pos" to start residue numbering there; default 1
  • Pretty rudimentary handling of PDB files, padding missing residues with '_' (i.e., no epitopes) and generally dealing only with the standard twenty 3-letter AA codes
  • In all cases, if the sequence is shorter than the epitope peptide length for that predictor (9 for Propred, 15 for NetMHCII), a score of 0 will be returned.

Predictors

The Predictors determine how to score the sequences.

  • --propred uses the Propred matrices to score the sequences. This is the default behaviour. It will look first in the current directory, then in the Rosetta database. It will attempt to look for the Rosetta database first using relative paths, assuming tools and main are cloned in the same directory. If that fails, if the environment variable $ROSETTA exists, it will assume main is cloned within that directory and will look for the database there.
  • --matrix MATRIX uses the MATRIX file to score the sequences.
  • --netmhcii uses the netmhcII executable, pointed to by the environment variable $NMHOME, to score the sequences.
  • --db DB uses a pre-computed SQL database, as generated using mhc_gen_db.py, to score the sequences.
  • --csv CSV uses a pre-computed CSV file, as generated using mhc_gen_db.py, to score the sequences.

Scoring Details

  • --allele_set SET tells the Predictor to use SET as the set of alleles.
    • With Propred, only all, southwood98, or test are allowed. If using test, only the DRB1_0101 allele will be used. The default (--allele_set all) is to use all alleles specified in the matrix.
    • With NetMHCII, there are four options: test will score only the DRB1_0101 allele, and greenbaum11 (27 alleles), southwood98 (8 alleles) and paul15 (7 alleles) will use published lists of alleles. all will use all 61 available alleles. test is the default. More alleles takes longer (not a huge issue with a single protein sequence), but may be more accurate.
  • --alleles ALLELE_LIST allows you to provide a custom list of alleles to use. Not used by default.
  • --epi_thresh EPI_THRESH sets the threshold of what constitutes a "hit." By default, this is 5.00, which means that the top 5% of binders are considered hits.
  • --noncanon {error,warn,silent} allows you to specify the behaviour if a residue other than the 20 canonical AAs is encountered (for example, ligands, non-canonical amino acids, etc.). Options are error (script will exit), warn (a warning will be printed), or silent. HOW ARE NCAAS DEALT WITH HERE? AS CHAIN BREAKS? WHAT IS THE DEFAULT BEHAVIOUR?
  • --netmhcii_score {rank,absolute} specifies whether to use the ranked score (i.e. is this in the top X% of binders) or absolute score (i.e. the affinity of binding) to determine if something is a hit. This will have an impact on the --epi_thresh parameter. By default, rank is used.
  • --db_unseen {error,warn,score} specifies, in a database predictor, what to do if a peptide that is absent from the database is encountered. Options are error (script will exit), warn (a warning will be printed, and a score of 0 will be assigned to that peptide), or score (the peptide will be scored using --db_unseen_score). warn is the default. IS THIS AN ACCURATE DESCRIPTION OF THE BEHAVIOUR?
    • --db_unseen_score SCORE specifies the penalty to apply if we encounter an "unseen" peptide and we are using --db_unseen score. A penalty of 100 is the default.

Output

The scoring results can be output in various ways. In any case where a filename is specified, using the $ character will substitute the sequence name in cases where multiple sequences are being scored.

  • --report {total,hits,full} specifies what kind of report to output to standard output. total will provide the total score for each sequence. hits will provide a report by allele for each peptide that is identified as a hit (i.e. score > 0). full will provide a report by allele for all peptides in the sequence. Default is total.
    • --report_file FILE will output the report generated by --report to a file in CSV format. If using multiple sequences, the $ symbol in the filename will be substituted with the sequence name or number.
  • --plot_hits_file FILE will output the results to a graphical file format. matplotlib must be installed for this to work. Plot format will be determined by the filename extension given.

Examples

Add some examples, probably from the demos.

mhc_gen_db.py

The mhc_gen_db.py script uses a lot of the same machinery as mhc_score.py, and so many of the same options work the same way as described above.

Input sequences

The input options work in the same way as mhc_score.py, but only one sequence can be used. This means the only available options are as follows:

  • --fa
  • --pdb/--chain

You may not specify a multi-FASTA file (--fsa), a peptides file (--pep), or sequences from standard input or the command line.

The option --firstres allows you to specify the residue number of the first residue in your PDB file. In cases where you have a PSSM based on PDB sequence (that doesn't necessarily start with residue 1), you will need this to correct the offset in the chain.

Controlling design space

Inherent in making a database of scores is the need to determine what your design space should be. This is a complex question to do the combinatorial aspects of sliding window peptides, as outlined in detail here. Because it is impossible to score all of design space even in a very small region, we provide a number of ways of describing which amino acids to allow at each position, and which positions to sample.

Note that regardless of these options, the entire input sequence will be scored and stored in the database.

Amino acid diversity

You may use one (and only one) of the following ways of specifying what amino acids to try at each position.

  • --aa_csv CSV_FILE specifies a CSV file where each line includes a position number followed by all allowed amino acids (one-letter code) at each position, each separated by commas. (Wild-type is automatically added in if missing.) See tools/mhc_energy_tools/demo/in/2sak_A.region_muts.csv for an example.
  • --pssm PSSM_FILE specifies a PSSM to use. Two PSSM formats are officially supported, though the script will try to parse unsupported formats as well. The two formats are the output from the web and command line versions of PSIBLAST. For more details on how to make a PSSM, see here. For examples of supported PSSMs, see tools/mhc_energy_tools/pssm_examples.
    • --pssm_thresh PSSM_THRESH sets the PSSM threshold above which a residue will be allowed. The PSSM should contain the log (base 2) of the observed substitution frequency at a given position divided by the expected substitution frequency at that position. A score > 1 means that the residue is observed more often than would be expected at random. The default is 1, though this can lead to very large database sizes if not increased.

Positions to sample

There are two options to limit which positions to sample. This is useful if, for example, you have a PSSM for the entire protein, but only want to score epitopes in specific regions.

  • --positions allows you to specify which regions to allow to mutate. Anything not included will not be scores (except the wild-type sequence). It should be formatted as follows: --positions 3-30 to target a single region (residues 3-30, inclusive) and --positions 3-30,113-142 to target multiple regions (residues 3-30 and 113-142, inclusive).
  • --lock prevents regions from mutating. It is specified in the same format as --positions.
  • If both --positions and --lock are specified, --lock overrides --positions. A lock position will not be mutated even if specified by --positions.

Predictors

All predictors available in mhc_score.py are available with mhc_gen_db.py, and are configured in the same way. Specifically:

  • --propred
  • --netmhcii
  • --matrix

In addition, the --netmhcii_raw NETMHC_OUTPUT option allows the raw output from a NetMHCII run to be used as input.

You may also use a SQL DB or CSV file generated by this script as your predictor. To do so, provide the path to said database/CSV file with the --db_in or --csv_in options.

Scoring Details

All of the options from mhc_score.py are available with mhc_gen_db.py to configure the predictors, except --db_unseen. Specifically:

  • --allele_set
  • --alleles
  • --epi_thresh
  • --noncanon
  • --netmhcii_score (for NetMHCII only)

Output

The database will be output to a file specified by either the --db or --csv options. If the file already exists at that location, the script will normally fail. You can specify to add to the existing database by providing the --augment argument for SQL DBs only. If you do so, it will verify that the database is using the same type of predictor and same list of alleles. If so, it will add any peptides that are missing from that database to the database. For either SQL DBs or CSV DBs, you may specify the --overwrite command to delete the existing file and start from scratch.

If no database is given, the script will run, following instructions from the additional output options, without scoring or saving the database.

Additional output options:

  • --estimate_size will calculate the number of peptides to be scored as configured. This is a good check to do before actually creating your database to get an idea of how big it will be. You can get a "back-of-napkin" estimate of the size of a SQL database by multiplying the number of peptides by 100, giving you an approximate file size in bytes for a 7 allele database. A database containing 3 million peptides will be about 300 MB, for example. The CSV file database will be about half the size.
  • --peps_out PEPTIDE.TXT will save all of the peptides to be scored in PEPTIDES.TXT, with one peptide per line.
  • --res_out RESFILE.RES will generate a resfile based on the design space specified. Only positions with more than one allowed amino acid will be listed in the resfile, using PIKAA to list them. Recall that wild-type is always allowed, so this is listing PIKAA lines for all positions that allow at least one non-native amino acid to be sampled.
    • --chain X, in addition to specifying what chain to use when a PDB file is used as input, will decide what chain ID to use in the resfile if a FASTA file is used as input. If input is using --fa, --chain is mandatory.
    • --res_header COMMAND applies COMMAND to the resfile header, above the start keyword, which will apply that command to all unspecified residues. By default, no header is included.
    • If you are using a PSSM to specify amino acid diversity and this PSSM was generated from a PDB file sequence (instead of from the sequence as translated), it is possible that the PSSM incorrectly assumes that the first residue is residue 1. If this is NOT the case, the --firstres parameters can be used to specify the real number of the first residue, allowing the PSSM and PDB sequence numbering to match up.

Multiprocessing

Because NetMHCII is slow compared to matrix-based scoring, spreading the computation for NetMHCII databases over multiple processors is enabled. Use the following commands to control multiprocessor database generation:

  • --nproc # determines the number of processors to use. By default, it uses 1 (no multiprocessing).
  • --batch # determines the number of peptides to pass to each processor at a time. NetMHCII will score that batch, and then another batch will be submitted to that processor until all peptides are scored.

Examples

Add some examples, probably from the demos.

mhc_data_db.py

The mhc_data_db.py script is similar to mhc_gen_db.py, but uses the IEDB database of experimentally-validated epitopes to construct the database. The user must download the IEDB to a local mysql server (can be done with this script) or as CSV files and provided as input.

Downloading the IEDB database

If you have a working mysql server, the easiest way of obtaining the IEDB is to download it using this script. You can do this using the following options:

  • --iedb_fresh_mysql DATABASE_NAME indicates that you want to download a fresh copy of the IEDB. DATABASE_NAME is a local mysql database that you can write to. Note that if this database exists, it will be overwritten! By default, "iedb" will be used as the database name.
  • --mysql_user USER and --mysql_pw PASSWORD are the username and password needed to access the local mysql database. By default, these will be "root" and no password.

Alternatively, you can download the IEDB as CSV files from their website.

Accessing the IEDB database

When generating your database/CSV file to use with Rosetta, you must specify how to access the IEDB.

If you are using a local mysql database, provide the following options:

  • iedb_mysql DATABASE_NAME gives the name of the local mysql database (downloaded as described above).
  • --mysql_user USER and --mysql_pw PASSWORD are the username and password needed to access the local mysql database. By default, these will be "root" and no password.

If you are using an IEDB CSV file, provide the following option:

  • iedb_csv CSV_FILE, where CSV_FILE is the name of the IEDB CSV file.

What data to use

You may specify what alleles to use. Four allele sets are supported: test, greenbaum11, paul15, southwood98, and hlaII. These can be selected using the --allele_set option. Alternatively, individual alleles can be specified using the --alleles option (only if not using a predictor, see below).

IEDB has MHC ligand binding data, and also MHC ligand elution data. If you are using a local mysql database, you can specify to use one, the other, or both:

  • --assay_mhc_ligand_binding controls whether to include the ligand binding data. Can be all to use, or none to leave out.
  • --assay_mhc_ligand_elution controls whether to include the ligand binding data. Can be all to use, or none to leave out.

If you are using an IEDB CSV file, you must curate the CSV appropriately yourself.

In addition, the IEDB contains data with peptide of arbitrary length. To work with mhc_epitope, we must reduce these to an appropriate 9mer size. This can be done in one of three ways:

  • --cores all will add all 9mers from the peptide. For example, if the peptide is an 11mer, three entries will be added to the database: one ranging from positions 1-9, one from 2-10, and one from 3-11.
  • --cores predicted_good means that any 9mer that beats a particular threshold, according to a predictor (see below) will be added to the database.
  • --cores predicted_best is not currently implemented, but will add the best 9mer (strongest binder), according to a predictor (see below).
  • --cores full will put the entire peptide, regardless of its length, into the database. This behaviour is not supported in Rosetta, but might be useful for other applications.

If you choose to select cores (i.e. predicted_good or predicted_best), you need to specify a predictor to score/rank them. Currently, you may use the following:

  • --propred to use Propred.
  • --pred_csv CSV_FILE to use a precomputed CSV file generated by this script or mhc_gen_db.py.
  • --pred_db SQL_DB_FILE to use a precomputed SQL DB file generated by this script or mhc_gen_db.py.
  • --matrix MATRIX_FILE to use a generic scoring matrix.

Note that the same allele_set specified for the IEDB will be used with your predictor.

In addition, you may pass options to control the predictor:

  • --epi_thresh THRESH sets the predictor's threshold for counting a peptide as a binder. Defaults to 5.0.
  • --noncanon sets how to deal with noncanonical residues.

Output

The database will be output to a file specified by either the --db or --csv options. If the file already exists at that location, the script will normally fail. You can specify to add to the existing database by providing the --augment argument for SQL DBs only. If you do so, it will verify that the database is using the same type of predictor and same list of alleles. If so, it will add any peptides that are missing from that database to the database. For either SQL DBs or CSV DBs, you may specify the --overwrite command to delete the existing file and start from scratch.

Examples

Add some examples, probably from the demos.

Using NetMHCII

To do (installation and setting environment)

List of alleles

The following is the complete list of alleles supported for each Predictor. In brackets, which allele_sets it occurs in is indicated (test, paul15, greenbaum11).

Propred

  • DRB1_0101 (test)
  • DRB1_0301
  • DRB1_0401
  • DRB1_0701
  • DRB1_0801
  • DRB1_1101
  • DRB1_1301
  • DRB1_1501

NetMHCII

  • DRB1_0101 (test, greenbaum11)
  • DRB1_0103
  • DRB1_0301 (paul15, greenbaum11)
  • DRB1_0401 (greenbaum11)
  • DRB1_0402
  • DRB1_0403
  • DRB1_0404
  • DRB1_0405 (greenbaum11)
  • DRB1_0701 (paul15, greenbaum11)
  • DRB1_0801
  • DRB1_0802 (greenbaum11)
  • DRB1_0901 (greenbaum11)
  • DRB1_1001
  • DRB1_1101 (greenbaum11)
  • DRB1_1201 (greenbaum11)
  • DRB1_1301
  • DRB1_1302 (greenbaum11)
  • DRB1_1501 (paul15, greenbaum11)
  • DRB1_1602
  • DRB3_0101 (paul15, greenbaum11)
  • DRB3_0202 (paul15, greenbaum11)
  • DRB3_0301
  • DRB4_0101 (paul15, greenbaum11)
  • DRB4_0103
  • DRB5_0101 (paul15, greenbaum11)
  • H-2-IAb
  • H-2-IAd
  • H-2-IAk
  • H-2-IAs
  • H-2-IAu
  • H-2-IEd
  • H-2-IEk
  • HLA-DPA10103-DPB10201 (greenbaum11)
  • HLA-DPA10103-DPB10301
  • HLA-DPA10103-DPB10401 (greenbaum11)
  • HLA-DPA10103-DPB10402
  • HLA-DPA10103-DPB10601
  • HLA-DPA10201-DPB10101 (greenbaum11)
  • HLA-DPA10201-DPB10501 (greenbaum11)
  • HLA-DPA10201-DPB11401 (greenbaum11)
  • HLA-DPA10301-DPB10402 (greenbaum11)
  • HLA-DQA10101-DQB10501 (greenbaum11)
  • HLA-DQA10102-DQB10501
  • HLA-DQA10102-DQB10502
  • HLA-DQA10102-DQB10602 (greenbaum11)
  • HLA-DQA10103-DQB10603
  • HLA-DQA10104-DQB10503
  • HLA-DQA10201-DQB10202
  • HLA-DQA10201-DQB10301
  • HLA-DQA10201-DQB10303
  • HLA-DQA10201-DQB10402
  • HLA-DQA10301-DQB10301
  • HLA-DQA10301-DQB10302 (greenbaum11)
  • HLA-DQA10303-DQB10402
  • HLA-DQA10401-DQB10402 (greenbaum11)
  • HLA-DQA10501-DQB10201 (greenbaum11)
  • HLA-DQA10501-DQB10301 (greenbaum11)
  • HLA-DQA10501-DQB10302
  • HLA-DQA10501-DQB10303
  • HLA-DQA10501-DQB10402
  • HLA-DQA10601-DQB10402

See Also