Category Archives: Python

How to build a Python dictionary of residues for each molecule in PyMOL

Sometimes it can be handy to work with multiple structures in PyMOL using Python.

Here’s a snippet of code you might find useful: we iterate over all the α-carbon atoms in a protein and append to a list tuples such as (‘GLY’, 1). The dictionary, ‘reslist’, returns a list of residue names and indices for each molecule, where the key is a string containing the name of the molecule. 

from pymol import cmd

# Create a list of all the objects, called 'mpls':
mols = cmd.get_object_list('*')

# Create an empty dictionary that will return a list of residues
# given the name of the molecule object
reslist = {}

# Set the dictionaries to be empty lists
for m in mols:  reslist[m] = []

# Use PyMOL's iterate command to go over every α-Carbon and append 
# a tuple consisting of the each residue's residue name ('resn') and
# residue index ('resi '):
for m in mols:  cmd.iterate('%s and n. ca'%m, 'reslist["%s"].append((resn,int(resi)))'%m)

This script assumes you only have protein molecules loaded, and ignores things like chain ID and insertion codes.

Once you have your list of residues, you can use it with the cmd.align command, e.g., to align a particular residue to a reference structure.

Using Conda environments with Flask and Apache

With the advent of ABlooper, we’ve recently introduced OpenMM as a new dependency for the SAbDab-SAbPred antibody modelling platform. By far the easiest way to install the OpenMM Python API is via Conda, so we’ve moved to Conda environments for the entire platform. This has made installation of the platform much easier, but introduces complications when it comes to running its web applications under Apache. In this post, I’ll briefly explain the reason for this, and provide a basic guide for running Flask apps using Conda environments under Apache.

Continue reading →

Automatic argument parsers for python

One of the recurrent problems I used to have when writing argument parsers is that after refactoring code, I also had to change the argument parser options which generally led to inconsistency between the arguments of the function and some of the options of the argument parser. The following example can illustrate the problem:

def main(a,b):
  """
  This function adds together two numbers a and b
  param a: first number
  param b: second number
  """
  print(a+b)

if __name__ == "__main__":
  import argparse
  parser = argparse.ArgumentParser()
  parser.add_argument("--a", type=int, required=True, help="first number")
  parser.add_argument("--b", type=int, required=True, help="second number")
  args = parser.parse_args()
  main(**vars(args))

This code is nothing but a simple function that prints a+b and the argument parser asks for a and b. The perhaps not so obvious part is the invocation of the function in which we have ** and vars. vars converts the named tuple args to a dictionary of the form {“a":1, "b":2}, and ** expands the dictionary to be used as arguments for the function. So if you have main(**{"a":1, "b":2}) it is equivalent to main(a=1, b=2).

Let’s refactor the function so that we change the name of the argument a to num.

Continue reading →

Meeko: Docking straight from SMILES string

When docking, using software like AutoDock Vina, you must prepare your ligand by protonating the molecule, generating 3D coordinates, and converting it to a specific file format (in the case of Vina, PDBQT). Docking software typically needs the protein and ligand file inputs to be written on disk. This is limiting as generating 10,000s of files for a large virtual screen can be annoying and hinder the speed at which you dock.

Fortunately, the Forli group in Scripps Research have developed a Python package, Meeko, to prepare ligands directly from SMILES or other molecule formats for docking to AutoDock 4 or Vina, without writing any files to disk. This means you can dock directly from a single file containing all the SMILES of the ligands you are investigating!

Continue reading →

Make your code do more, with less

When you wrangle data for a living, you start to wonder why everything takes so darn long. Through five years of introspection, I have come to conclude that two simple factors limit every computational project. One is, of course, your personal productivity. Your time of focused work, minus distractions (and yes, meetings figure here), times your energy and mental acuity. All those things you have little control over, unfortunately. But the second is the productivity of your code and tools. And this, in principle, is a variable that you have full control over.

Even quick calculations, when applied to tens of millions of sequences, can take quite some time!

This is a post about how to increase your productivity, by helping you navigate all those instances when the progress bar does not seem to go fast enough. I want to discuss actionable tools to make your code run faster, and generate more results, with less effort, in less time. Instructions to tinker less and think more, so you can do the science that you truly want to be doing. And, above all, I want to give out advice that is so counter-intuitive that you should absolutely consider following it.

Continue reading →

How to prepare a molecule for RDKit

RDKit is very fussy when it comes to inputs in SDF format. Using the SDMolSupplier, we get a significant rate of failure even on curated datasets such as the PDBBind refined set. Pymol has no such scruples, and with that, I present a function which has proved invaluable to me over the course of my DPhil. For reasons I have never bothered to explore, using pymol to convert from sdf, into mol2 and back to sdf format again (adding in missing hydrogens along the way) will almost always make a molecule safe to import using RDKit:

from pathlib import Path
from pymol import cmd

def py_mollify(sdf, overwrite=False):
    """Use pymol to sanitise an SDF file for use in RDKit.

    Arguments:
        sdf: location of faulty sdf file
        overwrite: whether or not to overwrite the original sdf. If False,
            a new file will be written in the form <sdf_fname>_pymol.sdf
            
    Returns:
        Original sdf filename if overwrite == False, else the filename of the
        sanitised output.
    """
    sdf = Path(sdf).expanduser().resolve()
    mol2_fname = str(sdf).replace('.sdf', '_pymol.mol2')
    new_sdf_fname = sdf if overwrite else str(sdf).replace('.sdf', '_pymol.sdf')
    cmd.load(str(sdf))
    cmd.h_add('all')
    cmd.save(mol2_fname)
    cmd.reinitialize()
    cmd.load(mol2_fname)
    cmd.save(str(new_sdf_fname))
    return new_sdf_fname

How to turn a SMILES string into a molecular graph for Pytorch Geometric

Despite some of their technical issues, graph neural networks (GNNs) are quickly being adopted as one of the state-of-the-art methods for molecular property prediction. The differentiable extraction of molecular features from low-level molecular graphs has become a viable (although not always superior) alternative to classical molecular representation techniques such as Morgan fingerprints and molecular descriptor vectors.

But molecular data usually comes in the sequential form of labeled SMILES strings. It is not obvious for beginners how to optimally transform a SMILES string into a structured molecular graph object that can be used as an input for a GNN. In this post, we show how to convert a SMILES string into a molecular graph object which can subsequently be used for graph-based machine learning. We do so within the framework of Pytorch Geometric which currently is one of the best and most commonly used Python-based GNN-libraries.

We divide our task into three high-level steps:

  1. We define a function that maps an RDKit atom object to a suitable atom feature vector.
  2. We define a function that maps an RDKit bond object to a suitable bond feature vector.
  3. We define a function that takes as its input a list of SMILES strings and associated labels and then uses the functions from 1.) and 2.) to create a list of labeled Pytorch Geometric graph objects as its output.
Continue reading →

List comprehension: an elegant Python feature inspired by mathematical set theory

Even though I have now deeply entered into the fascinating world of statistical machine learning and computational chemistry, my original background is very much in pure mathematics. Having spent some of my intellectually formative years in this highly purified and abstract universe, I still love to think in terms of sets, ordered tuples and well-defined functions whenever I have the luxury of being able to do so. This might be why list comprehension is one of my favourite features in Python.

List comprehension allows you to efficiently map a function over a list using elegant notation inspired by mathematical set theory. Let us first consider a (mathematical) set

A := \{1, 3, 7 \}.

Continue reading →

Uniformly sampled 3D rotation matrices

It’s not as simple as you’d think.

If you want to skip the small talk, the code is at the bottom. Sampling 2D rotations uniformly is simple: rotate by an angle from the uniform distribution \theta \sim U(0, 2\pi). Extending this idea to 3D rotations, we could sample each of the three Euler angles from the same uniform distribution \phi, \theta, \psi \sim U(0, 2\pi). This, however, gives more probability density to transformations which are clustered towards the poles:

Sampling Euler angles uniformly does not give an even distribution across the sphere.

In Fast Random Rotation Matrices (James Avro, 1992), a method for uniform random 3D rotation matrices is outlined, the main steps being:

Continue reading →

Out-of-distribution generalisation and scaffold splitting in molecular property prediction

The ability to successfully apply previously acquired knowledge to novel and unfamiliar situations is one of the main hallmarks of successful learning and general intelligence. This capability to effectively generalise is amongst the most desirable properties a prediction model (or a mind, for that matter) can have.

In supervised machine learning, the standard way to evaluate the generalisation power of a prediction model for a given task is to randomly split the whole available data set X into two sets – a training set X_{\text{train}} and a test set X_{\text{test}}. The model is then subsequently trained on the examples in the training set X_{\text{train}} and afterwards its prediction abilities are measured on the untouched examples in the test set X_{\text{test}} via a suitable performance metric.

Since in this scenario the model has never seen any of the examples in X_{\text{test}} during training, its performance on X_{\text{test}} must be indicative of its performance on novel data X_{\text{new}} which it will encounter in the future. Right?

Continue reading →