pyrosetta helper functions

Some functions that I keep using over and over as a Python module.

Installation

pip3 install pyrosetta-help

Colabs

Here are some Colabs notebooks I have put together for AlphaFold2 analyses:

Run PyRosetta in Colab
Analysis of dimer output from ColabFold in Colab
Migrate a ligand in Colab
Add missing loops by cannibilising AlphaFold2
Stretch out an alphafold pose — ToDo
Add OPM dots to show membrane — ToDo

Cleaner Pip install for pyrosetta

Whereas one can create a conda environment easy fully loaded with fun, e.g.

CONDA_OVERRIDE_GLIBC=2.35 conda create -n 👾👾👾 python=3.8 -y \
        -c conda-forge \
        -c https://👾👾👾:👾👾👾@west.rosettacommons.org/pyrosetta/conda/release/ \
        -c schrodinger \
        -c plotly \
        pyrosetta pymol-bundle rdkit plotly dask

Were the alien emoji are redacting the username and password.

Installing with pip is more problematic. As of Febuary 2022 either of the following fail because the authentication on the 302 redirect causes issue:

pip install https://👾👾👾:👾👾👾@https://graylab.jhu.edu/download/PyRosetta4/archive/release/
pip install https://👾👾👾:👾👾👾@https://graylab.jhu.edu/download/PyRosetta4/archive/release/PyRosetta4.Release.python39.mac.wheel/latest.html

As a result pyrosetta_help has two functions, install_pyrosetta and check_pyrosetta, which aim to help this. The setup.py also registers a command (install_pyrosetta) to make it possible too.

pip install pyrosetta-help
install_pyrosetta -u 👾👾👾 -p 👾👾👾

In Python, the module pyrosetta_help needs to be reloaded afterwards.

import pyrosetta_help as ph  # this will give a warning because there's no pyrosetta
print(ph.check_pyrosetta())  # False, there is no pyrosetta
ph.install_pyrosetta('👾👾👾', '👾👾👾')
from importlib import reload
reload(ph)

No plain text username+password combinations are stored in this repository. But the inputted values are SHA256-hashed and compared to a hash of the correct Rosetta and PyRosetta credentials. If the former are used a clear error message will be shown as Rosetta has a different set of credentials.

Hit: The pyrosetta username and password are in the format like boltzmann + constant, not SomeThingUser+qwerty.

(This will however not stop people from asking, ae)

If the credentials are incorrect an error is raised, unless hash_comparison_required=False.

install_pyrosetta(username= username,
                  password= password,
                  path=None,
                  hash_comparison_required=True)

In the case of a ReadTheDocs confuguration, adding pyrosetta_help to the requirements.txt file and then adding in builds of the .readthedocs.yml file the following:

builds:
  ...
  jobs:
    post_install:
      - install_pyrosetta

And then adding in advanced options the private environment variables PYROSETTA_USERNAME and PYROSETTA_PASSWORD, will result in a sucessful build.

Starting up

A few helper functions.

get_logger, get_log_entries

The function configure_logger, simply adds a stringIO handler to the log and captures the log. The function get_log_entries, spits out entries of a given level.

make_option_string

This just converts the key:value pairs to a command line string for the pyrosetta init.

Bools are converted,
None results in a value argument,
Tuples are converted to xx:xx type arguments
Dictionaries are converted to xx:xx type arguments (multiple, if multiple keys in the nested dictionary)

import pyrosetta
import pyrosetta_help as ph

# capture to log
logger = ph.configure_logger()
# give CLI attributes in a civilised way
pyrosetta.distributed.maybe_init(extra_options=ph.make_option_string(no_optH=False,
                                                ex1=None,
                                                ex2=None,
                                                #mute='all',
                                                ignore_unrecognized_res=True,
                                                load_PDB_components=False,
                                                ignore_waters=False)
                               )
# ...
# show relevant error
print(ph.get_log_entries('ERROR'))

Common operations

Import a file, while dealing with the param files

import pyrosetta_help as ph
pose = ph.pose_from_file('combined.relaxed2.pdb', params_filenames=['35G.params','CMP.params', 'ATP.params', 'NME.params'])

I have somewhere one that via rdkit_to_params starts with a dict of residue:SMILES. TODO find.

Get pandas dataframe of score

import pyrosetta_help as ph
scores = ph.pose2pandas(pose)
scores.loc[scores.total_score > 10][['residue', 'total_score']]

Convert a selector to a list of str of NGL selector style [resn]resi:chain

import pyrosetta_help as ph
ligand = pyrosetta.rosetta.core.chemical.ResidueProperty.LIGAND
lig_sele = pyrosetta.rosetta.core.select.residue_selector.ResiduePropertySelector(ligand)
ph.clarify_selector(lig_sele, pose)

Local relax, etc.

import pyrosetta_help as ph
map_filename:str = ph.download_map('1A2A') # cryoEM require EMD codes
ed = ph.prep_ED(pose, map_filename)
local_scorefxn = ph.get_local_scorefxn()
local_relax = ph.get_local_relax()
local_relax.apply(pose)

Note, ph.add_bfactor_from_score is unstable!

score_mutants

Given a list of mutants and pose, score them. scorefunction terms, interface, movement etc.

import pyrosetta_help as ph

model = ph.MutantScorer(pose, modelname='test')
model.scorefxn = pyrosetta.create_score_function('ref2015')
model.strict_about_starting_residue = True
data = model.score_mutations(['p.Met1Gly', 'p.Ser29Glu'],
                             chain='V',
                             interfaces=(('isolated', 'V_ABCDEFGHIJKLMNOPQRSTWXYZ'),),  #
                             preminimize=True,
                             distance=12,
                             cycles=5)
import pandas as pd

scores = pd.DataFrame(data)
ph.extend_scores(scores)

The function extend_scores adds 6 columns, specifying which terms is the biggest changer.

Blueprinter

A key component of using Remodel is a blueprint. This module makes a blueprint. See doc string of class Blueprinter in blueprint_maker for more.

from pyrosetta_help import Blueprinter
blue = Blueprinter.from_pose(pose)
blue[20:25] = 'NATAA' # wobble
blue.wobble_span(20,25) # same as above.
del blue[15:20] # requires preceeding and suceeding residues to be NATAA though!
blue.del_span(15, 20) # same as above, but wobbles the preceeding and suceeding 1 residues
blue[22] = 'PIKAA W' # requires wobble
blue.mutate(22, 'W') # same as above, but wobbles the preceeding and suceeding residues

To set it:

blue.set('mut.blu')

This equivalent to the following (handy if something needs manual correction)

blue.write('mut.blu')
blue.bluprint = 'mut.blu'
# which calls `pyrosetta.rosetta.basic.options.set_file_option('remodel:blueprint', 'mut.blu')`
# so do not forget `mover.register_options()`!

This can therefore be used as normal:

pyrosetta.rosetta.basic.options.set_boolean_option('remodel:quick_and_dirty', True)
pyrosetta.rosetta.basic.options.set_string_option('remodel:generic_aa', 'G')

rm = pyrosetta.rosetta.protocols.forge.remodel.RemodelMover()
rm.register_options()
rm.dr_cycles(5) # default 3
rm.max_linear_chainbreak(0.2) # default 0.07
rm.redesign_loop_neighborhood(False)
rm.apply(pose)

WeightWatcher

A class to easily look at the weights for the terms of different scorefunctions. It has various functionalities.

from pyrosetta_help.weights import WeightWatcher
ww = WeightWatcher()
# List available scorefunctions:
print( ww.possible_scorefxn_names ) # dynamic attribute
# Find a scorefunction that mentions a word:
ww.find_metion('foo')
# Get the comment block of a scorefunction:
print(ww.get_scorefxn_comments('ref2015'))
# Get a scorefunction by name by calling the appropriate options first.
# NB. calling a different one will change the options.
scorefxn = ww.get_scorefxn('beta_nov16')
# Get weights dictionary (including ref) for a scorefunction or a name of one
weights = ww.get_weights('beta_nov16')
# Get a pandas table of the weights
weight_table = ww.compare(['ref2015', 'beta_nov16'], different_only=True)

It also has the class attribute term_meanings, which is a handy dictionary to convert a score term name into a description. E.g. converts "fa_atr" -> "Lennard-Jones attractive between atoms in different residues (r^6 term, London dispersion forces)." etc. Taken from Rosetta documentations, with some edits on some terms.

ChainOps

This class works around a list of dict that contain details of each chain in an model. The key keys are:

number (pose chain number)
chain (PDB chain letter)
gene_name (gene name)

the instance can be subscripted with any of those three, returning the dict of that chain.

from pyrosetta_help import ChainOps

To get a single chain pose:

chain_pose = chain_ops.get_pose_of_chain(pose, 'B')

Transmogrifer/Murinizer deals with alignments between species. The RosettaCM stuff is elsewhere. To do: MOVE OVER

The chain details started off by themselves, see metadata_assembly notes

Other snippets

These could quickly be made into classes... but hey

Legal disclaimer

I (Matteo Ferla) am not affiliated with PyRosetta or Rosetta Commons, I am just an avid user. My usage does not constitute an endorsement of PyRosetta by the BRC, NIHR, Wellcome Trust, the University of Oxford, the United Kingdom of Great Britain + Northern Ireland + all its dependencies etc. etc.