This is my first OPIG blog! I’m going to start with a summary of the Graphormer, a Graph Neural Network (GNN) that borrows concepts from Transformers to boost performance on graph tasks. This post is largely based on the NeurIPS paper Do Transformers Really Perform Bad for Graph Representation?by Ying et. al., which introduces the Graphormer, and which we read for our last deep learning journal club. The project has now been integrated as a Microsoft Research project.
I’ll start with a cheap and cheerful summary of Transformers and GNNs before diving into the changes in the Graphormer. Enjoy!
Graph neural networks (GNNs) have quickly become one of the most important tools in computational chemistry and molecular machine learning. GNNs are a type of deep learning architecture designed for the adaptive extraction of vectorial features directly from graph-shaped input data, such as low-level molecular graphs. The feature-extraction mechanism of most modern GNNs can be decomposed into two phases:
Message-passing: In this phase the node feature vectors of the graph are iteratively updated following a trainable local neighbourhood-aggregation scheme often referred to as message-passing. Each iteration delivers a set of updated node feature vectors which is then imagined to form a new “layer” on top of all the previous sets of node feature vectors.
Global graph pooling: After a sufficient number of layers has been computed, the updated node feature vectors are used to generate a single vectorial representation of the entire graph. This step is known as global graph readout or global graph pooling. Usually only the top layer (i.e. the final set of updated node feature vectors) is used for global graph pooling, but variations of this are possible that involve all computed graph layers and even the set of initial node feature vectors. Commonly employed global graph pooling strategies include taking the sum or the average of the node features in the top graph layer.
While a lot of research attention has been focused on designing novel and more powerful message-passing schemes for GNNs, the global graph pooling step has often been treated with relative neglect. As mentioned in my previous post on the issues of GNNs, I believe this to be problematic. Naive global pooling methods (such as simply summing up all final node feature vectors) can potentially form dangerous information bottlenecks within the neural graph learning pipeline. In the worst case, such information bottlenecks pose the risk of largely cancelling out the information signal delivered by the message-passing step, no matter how sophisticated the message-passing scheme.
A few weeks ago we attended the RSC’s 5th AI in Chemistry conference at Churchill College, Cambridge. It featured research and discussions on a broad range of topics and was attended by a diverse set of more than 200 researchers from academia and industry, including six Opiglets.
Recently I was tasked with filtering out ‘stringy’ molecules that were being produced with the fragment merging method I’m working on (that is, molecules with lots of consecutive non-ring bonds that weren’t necessarily caught with my rotatable bond filter). While this is quite a niche/specific task, through this I discovered a couple of RDKit functions that I wasn’t previously aware of but might be helpful for other people regularly looking at small molecules. The demo adapts code from this helpful blogpost on cutting a molecule into rings and linkers from ‘Is life worth living?’ (which is a useful source of cheminformatics wisdom; https://iwatobipen.wordpress.com/2020/01/23/cut-molecule-to-ring-and-linker-with-rdkit-rdkit-chemoinformatics-memo/). Obviously in practice you may be applying lots of different filters to enumerated molecules, but this is just a small example of something I found useful.
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!
The lineup for the Royal Society of Chemistry’s 5th “Artificial Intelligence in Chemistry” Symposium (Thursday-Friday, 1st-2nd September 2022) is now complete for both oral and poster presentations. It really is a fantastic selection of topics and speakers and it is clear this event is now a highlight of the scientific calendar. Our very own Prof. Charlotte M. Deane, MBE will be giving a keynote.
Finding a way to express the similarity of irregular and discrete molecular graphs to enable quantitative algorithmic reasoning in chemical space is a fundamental problem in data-driven small molecule drug discovery.
Virtually all algorithms that are widely and successfully used in this setting boil down to extracting and comparing (multi-)sets of subgraphs, differing only in the space of substructures they consider and the extent to which they are able to adapt to specific downstream applications.
A large body of recent work has explored approaches centred around graph neural networks (GNNs), which can often maximise both of these considerations. However, the subgraph-derived embeddings learned by these algorithms may not always perform well beyond the specific datasets they are trained on and for many generic or resource-constrained applications more traditional “non-parametric” topological fingerprints may still be a viable and often preferable choice .
This blog post gives an overview of the topological fingerprint algorithms implemented in RDKit. In general, they count the occurrences of a certain family of subgraphs in a given molecule and then represent this set/multiset as a bit/count vector, which can be compared to other fingerprints with the Jaccard/Dice similarity metric or further processed by other algorithms.
As a reasonably new RDKit user, I was relieved to find that using its built-in functionality for generating basic images from molecules is quite easy to use. However, over time I have picked up some additional tricks to make the images generated slightly more pleasing on the eye!
The first of these (which I definitely stole from another blog post at some point…) is to ask it to produce SVG images rather than png:
#ensure the molecule visualisation uses svg rather than png format
IPythonConsole.ipython_useSVG=True
Now for something slightly more interesting: as a fragment elaborator, I often need to look at a long list of elaborations that have been made to a starting fragment. As these have usually been docked, these don’t look particularly nice when loaded straight into RDKit and drawn:
#load several mols from a single sdf file using SDMolSupplier
#add these to a list
elabs = [mol for mol in Chem.SDMolSupplier('frag2/elabsTestNoRefine_Docked_0.sdf')]
#get list of ligand efficiencies so these can be displayed alongside the molecules
LEs = [(float(mol.GetProp('Gold.PLP.Fitness'))/mol.GetNumHeavyAtoms()) for mol in elabs]
Draw.MolsToGridImage(elabs, legends = [str(LE) for LE in LEs])
Fig. 1: Images generated without doing any tinkering
Two quick changes that will immediately make this image more useful are aligning the elaborations by a supplied substructure (here I supplied the original fragment so that it’s always in the same place) and calculating the 2D coordinates of the molecules so we don’t see the twisty business happening in the bottom right of Fig. 1:
Molecular descriptors are quantities associated with small molecules that specify physical or chemical properties of interest. They can be used to numerically describe many different aspects of a molecule such as:
molecular graph structure,
lipophilicity (logP),
molecular refractivity,
electrotopological state,
druglikeness,
fragment profile,
molecular charge,
molecular surface,
…
Vectors whose components are molecular descriptors can be used (amongst other things) as high-level feature representations for molecular machine learning. In my experience, molecular descriptor vectors tend to fall slightly short of more low-level molecular representation methods such as extended-connectivity fingerprints or graph neural networks when it comes to predictive performance on large and medium-sized molecular property prediction data sets. However, one advantage of molecular descriptor vectors is their interpretability; there is a reasonable chance that the meaning of a physicochemical descriptor can be intuitively understood by a chemical expert.
A wide variety of useful molecular descriptors can be automatically and easily computed via RDKit purely on the basis of the SMILES string of a molecule. Here is a code snippet to illustrate how this works:
