pIChemiSt

Description

The program calculates the isoelectric point of proteins or peptides based on their 2D molecular structure. The input structure is cut into monomers by targeting its amide bonds, and then each monomer's pKa value is determined using different methods: natural amino acids pKa values are matched against a dictionary; non-natural amino acid values are calculated using either pKaMatcher (built-in tool based on SMARTS patterns) or ACD perceptabat GALAS algorithm (a commercial tool that requires licence). For natural amino acids the following predefined sets of amino-acid pKa values are implemented: 'IPC2_peptide', 'IPC_peptide', 'ProMoST', 'Gauci', 'Rodwell', 'Grimsley', 'Thurlkill', 'Solomon', 'Lehninger', 'EMBOSS'. The mean value and variation between different sets are also calculated as well as the total charge at pH 7.4. The program can also plot the corresponding pH/Q curves for each input structure. Please refer to pIChemiSt publication for more details: https://pubs.acs.org/doi/10.1021/acs.jcim.2c01261

How to install the software via pypi

• Ensure that you have Python version >=3.9
• Run pip install pichemist
• (optional) - To use ACD for the prediction of non-natural amino acid pKa, make sure that the command perceptabat points to its binary

How to install the software via Github

• Clone the repository
• Ensure that you have Python version >=3.9
• Enter the package folder cd peptide-tools/pIChemiSt
• Run pip install . to install the Python library and the CLI command
• (optional) - To use ACD for the prediction of non-natural amino acid pKa, make sure that the command perceptabat points to its binary

Examples of usage (CLI)

# Run the predictor against a SMILES file using pKaMatcher and output results to console
pichemist -i test/examples/payload_1.smi --method pkamatcher

>
======================================================================================================================================================
pI
---------------------------------
     pI mean  9.02
         err  0.61
         std  1.72
IPC2_peptide  8.05
 IPC_peptide  9.81
     ProMoST  8.38
       Gauci  8.69
    Grimsley  8.94
   Thurlkill  9.06
   Lehninger  9.86
    Toseland  9.41


======================================================================================================================================================
Q at pH7.4
---------------------------------
Q at pH7.4 mean  0.73
         err  0.24
         std  0.67
IPC2_peptide  0.63
 IPC_peptide  0.99
     ProMoST  0.26
       Gauci  0.55
    Grimsley  0.66
   Thurlkill  0.8
   Lehninger  0.99
    Toseland  0.95


pH interval with charge between -0.2 and  0.2 and prediction tool: pkamatcher
pI interval:  8.6 -  9.4

Other flavours of flags and arguments can be configured to feed different inputs or produce different outputs:

# Use SMILES string as input
pichemist -i "N[C@@]([H])(CS)C(=O)N[C@@]([H])(CC(=O)N)C(=O)N[C@@]([H])(CS)C(=O)N[C@@]([H])(CC(=O)N)C(=O)O" -if smiles_stdin

# Use SMILES string as input and output JSON to console
pichemist -i "C([C@@H](C(=O)O)N)SSC[C@@H](C(=O)O)N" -if smiles_stdin -of json

# Use FASTA as input
# Note that FASTA assumes that C- and N- termini are ionisable by default
pichemist -i "MNSERSDVTLY" -if fasta_stdin

# Use FASTA with capped termini, i.e., non-ionisable, for both C- and N-.
# These can be configured as preferred by removing the corresponding flags.
# This configuration can also be used as a 'trick' for feeding cyclic peptides
# as linear FASTA sequences as their termini will be assumed to be non ionisable.
pichemist -i "MNSERSDVTLY" -if fasta_stdin --ionizable_nterm false --ionizable_cterm false

# Use SDF as input
pichemist -i test/examples/payload_4.sdf -if sdf

# Output SDF
pichemist -i test/examples/payload_1.smi -o results.sdf -of sdf

# Output JSON to console
pichemist -i test/examples/payload_1.smi -of json

# Plot pH/Q curve
pichemist -i test/examples/payload_1.smi --plot_ph_q_curve

# Plot pH/Q curve (with custom prefix 'plot')
pichemist -i test/examples/payload_2.smi --plot_ph_q_curve -pp "plot"

# Print the pKa values of fragments
pichemist -i test/examples/payload_1.smi --print_fragment_pkas

# Use ACD instead of pKaMatcher
pichemist -i test/examples/payload_1.smi --method acd

# Use ACD with SMILES string
pichemist -i "NCCC(=O)N[C@@H](Cc1c[nH]cn1)C(=O)O" --method acd -if smiles_stdin

Examples of usage (Python API)

import pprint
from pichemist.io import generate_input
from pichemist.api import pichemist_from_dict

smiles = "C[C@@H](NC(=O)[C@H](CCCCN)NC(=O)[C@](C)(CC(=O)O)NC(=O)[C@H](CCCN)NC(=O)[C@@H](N)Cc1ccccc1)C(=O)O"

args = {
        "input_data": smiles,
        "input_format": "smiles_stdin",
        "plot_ph_q_curve": False,
        "print_fragments": False,
        "method": "pkamatcher",
    }

input_dict = generate_input(args["input_format"], args["input_data"])
output = pichemist_from_dict(
    input_dict, args["method"], args["plot_ph_q_curve"], args["print_fragments"]
)

pp = pprint.PrettyPrinter(depth=4)
pp.pprint(output)
>
{1: {'QpH7': {'Gauci': 0.5541,
              'Grimsley': 0.6645,
              'IPC2_peptide': 0.6315,
              'IPC_peptide': 0.9916,
              'Lehninger': 0.9932,
              'ProMoST': 0.2617,
              'Q at pH7.4 mean': 0.7307,
              'Thurlkill': 0.7975,
              'Toseland': 0.9516,
              'err': 0.2386,
              'std': 0.675},
     'mol_name': 'C[C@@H](NC(=O)[C@H](CCCCN)NC(=O)[C@](C)(CC(=O)O)NC(=O)[C@H](CCCN)NC(=O)[C@@H](N)Cc1ccccc1)C(=O)O',
     'pI': {'Gauci': 8.6875,
            'Grimsley': 8.9375,
            'IPC2_peptide': 8.046875,
            'IPC_peptide': 9.8125,
            'Lehninger': 9.859375,
            'ProMoST': 8.375,
            'Thurlkill': 9.0625,
            'Toseland': 9.40625,
            'err': 0.6087,
            'pI mean': 9.0234,
            'std': 1.7216},
     'pI_interval': (8.625, 9.3625),
     'pI_interval_threshold': 0.2,
     'pKa_set': 'IPC2_peptide'}}

Contributions

• Andrey I. Frolov (https://orcid.org/0000-0001-9801-3253)
• Gian Marco Ghiandoni (https://orcid.org/0000-0002-2592-2939)
• Jonas Boström - contributed the function of splitting molecules by peptide bonds (https://orcid.org/0000-0002-9719-9137)
• Johan Ulander - contributed the idea to do SMARTS pattern matching Smiles into Amino Acid single letter code (https://orcid.org/0009-0004-7655-2212)

For developers

• To create a new build, the package version first needs to be configured inside setup.py then the command python setup.py sdist will build the distribution. The command bdist_wheel should not be used since this mode in setup.py skips including the required static files in the wheel distribution
• The code can be automatically tested using python setup.py test which requires pytest to be installed
• Tests can also be run using the Makefile in the root of the repository. The file allows granular testing as follows:
  • make test_core runs only the core tests including pKaMatcher and plots
  • make test_acd only runs acd tests (which require an ACD license)
  • make test runs both core and acd tests
• We strongly recommend using pre-commit when contributing to this repo. The root folder of peptide-tools contains a .pre-commit-config.yaml which can be used to set a pre-commit hook and automatically run a series of validations.