settmhc 1.0.0

A peptide-MHC binding predictor based on sequence-structure information

SETTMHC

SETTMHC is a peptide-MHC class I binding predictor software based on multimodal sequence-structure information.

SETTMHC also includes other two modules: pMHC complex structure mdeling and a persistent homology method to analysis pMHC complex surface topology features as structure representations.

Installation

System

Linux

One-step installation

Befor using the environment.yaml, make sure the Pyrosetta download channel is modified in the file.

Using conda, just run conda env create -n SETTMHC -f environment.yaml, but it may cause some dependencies issues.

Build step by step

Due to the two structure modeling methods, the environment build may be cumbersome.

python > 3.7

1. Requirements for APE-Gen

1.1 Conda packages

• conda install -c bioconda smina
• conda install -c omnia pdbfixer
• conda install -c conda-forge mdtraj
• conda install -c schrodinger pymol or conda install -c conda-forge pymol-open-source
• conda install -c bioconda autodock-vina
• conda install -c conda-forge openmm=7.5.1

1.2 Install RCD

• Download RCD v1.4 from https://chaconlab.org/modeling/rcd/rcd-download
• Add xxxx/RCD_v1.40_Linux_20190228/bin to PATH to make 'rcd' command available
• Make sure the lib path containing 'libmkl_intel_lp64.so' is added to LD_LIBRARY_PATH, you may get the lib path by find /home -name libmkl_intel_lp64.so. If there is no such object, run conda install -c intel mkl and add the lib path to LD_LIBRARY_PATH.

2. Requirements for RosettaMHC

• Download Pyrosetta
• conda install biopython
• Download Clustal omega and add it to PATH

3. Install SETTMHC from pip

Before install settmhc, you may install pytorch 1.9.0 (cpu only or with cuda) manually depending on your device.

pip install settmhc

Usage

SETTMHC binding prediction for pMHC

Download the model files first, defaluts to ~/.local/share/settmhc/, or you can add export SETTMHC_MODEL_DIR="path you like" to your ~/.bashrc beforehand.

settmhc --download

You can run prediction with peptides and hlas input:

settmhc --peptides KELEGILLL ALLGLTLGV --hlas HLA-A*02:01 HLA-A*11:01 -o output.csv

Or just input a csv file contains HLA and peptides, for more details in input.csv, see demo/demo.csv, which Annotation column is optional.

settmhc -i input.csv -o output.csv

Or specify a pMHC complex PDB file

settmhc -pdb 1DUZ.pdb --protein A --peptide C

See settmhc -h for more argument information.

pMHC complex structre modeling

usage: pmhc-model [-h] [--software {APE-Gen,RosettaMHC}] [--output_dir OUTPUT_DIR] [--filename FILENAME]
                  [--list_supported_hlas] [--template_pdb TEMPLATE_PDB] [--mhc_chain MHC_CHAIN]
                  [--peptide_chain PEPTIDE_CHAIN]
                  peptide_input receptor_class

Peptide-MHC complex structure Generator

positional arguments:
  peptide_input         Sequence of peptide
  receptor_class        MHC class I allele

optional arguments:
  -h, --help            show this help message and exit
  --software {APE-Gen,RosettaMHC}
                        Software used for pMHC structure generation, defaults to 'APE-Gen' (default: APE-Gen)
  --output_dir OUTPUT_DIR
                        Output PDB file saved path (default: structures)
  --filename FILENAME   Saved PDB file name,e.g.,default='pmhc001', and the output file will be like 'pmhc001.pdb'
                        (default: pmhc001)
  --list_supported_hlas
                        List all the supported 110 HLA alleles (default: False)

RosettaMHC simple arguments:
  --template_pdb TEMPLATE_PDB
                        Provide template structure in PDB to perform threading (default: None)
  --mhc_chain MHC_CHAIN
                        HLA chain in the template PDB file (default: A)
  --peptide_chain PEPTIDE_CHAIN
                        Peptide chain in the template PDB file (default: C)

To model pMHC structure, only input one pair of peptide and HLA allele at a time.

pmhc-model KELEGILLL HLA-A*02:01 

Persistent homology method for structure analysis

usage: pshm [-h] -i INPUT [-o OUTPUT] [-l LENGTH] [-f FILTRATION] [-u UNIT] [-c CUTOFF] [--protein PROTEIN]
            [--ligand LIGAND]

A persistent homology method for pMHC interface structure feature generation

optional arguments:
  -h, --help            show this help message and exit
  -i INPUT, --input INPUT
                        Input file for structure feature generation (default: None)
  -o OUTPUT, --output OUTPUT
                        Output file for pshm results containing the 16×360 feature matrix (default: None)
  -l LENGTH, --length LENGTH
                        Max length for RipsComplex (default: 15)
  -f FILTRATION, --filtration FILTRATION
                        Max length for filtration process, usually the same as RipsComplex max length (default: 15)
  -u UNIT, --unit UNIT  Filtration unit (length for each filtration shell) (default: 0.5)
  -c CUTOFF, --cutoff CUTOFF
                        HLA and peptide atom distance cutoff for HLA atoms filter (default: 10)
  --protein PROTEIN     HLA chain in the PDB file (default: A)
  --ligand LIGAND       Peptide chain in the PDB file (default: C)

The default arguments for RipsComplex building are used for SETTMHC training.

pshm -i 1DUZ.pdb -o 1duz_features.csv --protein A --ligand -C

Potential issues

An error may occur during the application AttributeError: module 'distutils' has no attribute 'version' due to the high version of setuptools.

You can fix this by pip install setuptools==59.5.0