At OPIG Group Meetings, it’s customary to give “Conference Feedback” whenever any of us has recently attended a conference. Typically, people highlight the most interesting talks—either to them or others in the group.
Having just returned from the 6th RSC-BMCS / RSC-CICAG AI in Chemistry Symposium, it was my turn last week. But instead of the usual perspective—of an attendee—I spoke briefly about how to organize a conference.
Continue readingWhen designing machine learning algorithms with the aim of accelerating the discovery of novel and more effective therapeutics, we often care deeply about their ability to generalise to new regions of chemical space and accurately predict the properties of molecules that are structurally or functionally dissimilar to the ones we have already explored. To evaluate the performance of algorithms in such an out-of-distribution setting, it is essential that we are able to quantify the data shift that is induced by the train-test splits that we rely on to decide which model to deploy in production.
For our recent ICML 2023 paper Drug Discovery under Covariate Shift with Domain-Informed Prior Distributions over Functions, we chose to quantify the distributional similarity between two sets of molecules through the Maximum Mean Discrepancy (MMD).
Continue readingWe’ve been investigating deep learning-based protein-ligand docking methods which often claim to be able to generate ligand binding modes within 2Å RMSD of the experimental one. We found, however, this simple criterion can conceal a multitude of chemical and structural sins…
DeepDock attempted to generate the ligand binding mode from PDB ID 1t9b
(light blue carbons, left), but gave pretzeled rings instead (white carbons, right).
When showcasing an approach in computational chemistry, an example molecule is required as a placeholder. But which to chose from? I would classify there different approaches: choosing a recognisable molecules, a top selling drugs, or a randomly sketched compound.
At a recent conference, Sheffield Cheminformatics 2023, I saw examples of all three and one problem I had that some placeholders distracted me into searching to figure out what it was.
Continue readingOver the next few days, researchers from around the world will be gathering in Sheffield for the 9th Joint Sheffield Conference on Cheminformatics. As one of the organizers (wearing my Molecular Graphics and Modeling Society ‘hat’), I can say we have an exciting array of speakers and sessions:
It has traditionally taken place every three years, but despite the global pandemic it is returning this year, once again in person in the excellent conference facilities at The Edge. You can download the full programme in iCal format, and here is the conference calendar:
Continue readingFinding the parts in common between two molecules appears to be a straightforward, but actually is a maze of layers. The task, maximum common substructure (MCS) searching, in RDKit is done by Chem.rdFMCS.FindMCS, which is highly customisable with lots of presets. What if one wanted to control in minute detail if a given atom X and is a match for atom Y? There is a way and this is how.
I thought I’d share a list of some of the other blogs that have helped me during my PhD so far and may be useful to new starters (or those who may not have come across them before). This list is by no means exhaustive and I’m very open to other recommendations!
Continue reading
Deep learning (DL) methods in structural modelling are outcompeting force fields because they overcome the two main limitations to force fields methods – the prohibitively large search space for large systems and the limited accuracy of the description of the physics [4].
However, the two methods are also compatible. DL methods are helping to close the gap between the applications of force fields and ab initio methods [3]. The advantage of DL-based force fields is that the functional form does not have to be specified explicitly and much more accurate. Say goodbye to the 12-6 potential function.
In principle DL-based force fields can be applied anywhere where regular force fields have been applied, for example conformation generation [2]. The flip-side of DL-based methods commonly is poor generalization but it seems that force fields, when properly trained, generalize well. ANI trained on molecules with up to 8 heavy atoms is able to generalize to molecules with up to 54 atoms [1]. Excitingly for my research, ANI-2 [2] can replace UFF or MMFF as the energy minimization step for conformation generation in RDKit [5].
So let’s use Auto3D [2] to generated low energy conformations for the four molecules caffeine, Ibuprofen, an experimental hybrid peptide, and Imatinib:
CN1C=NC2=C1C(=O)N(C(=O)N2C)C CFF CC(C)Cc1ccc(cc1)C(C)C(O)=O IBP Cc1ccccc1CNC(=O)[C@@H]2C(SCN2C(=O)[C@H]([C@H](Cc3ccccc3)NC(=O)c4cccc(c4C)O)O)(C)C JE2 Cc1ccc(cc1Nc2nccc(n2)c3cccnc3)NC(=O)c4ccc(cc4)CN5CCN(CC5)C STIContinue reading
Recently I’ve been thinking a lot about how to decompose a compound into smaller fragments specifically for a retrosynthetic purpose. My question is: given a compound, can I return building blocks that are likely to synthesize together to produce this compound simply by breaking likely bonds formed in a reaction? A method that is nearly 15 years old named, breaking of retrosynthetically interesting chemical substructures (BRICS), is one approach to do this. Here I’ll explore how BRICS can reflect synthetic accessibility.
Continue readingWhen recently looking at some reaction data, I was confronted with the problem of atom-to-atom mapping (AAM) and what tools are available to tackle it. AAM refers to the process of mapping individual atoms in reactants to their corresponding atoms in the products, which is important for defining a reaction template and identifying which bonds are being formed and broken. This has many downstream uses for computational chemists, such as for reaction searching and forward and retrosynthesis planning1. The problem is that many reaction databases do not contain these mappings, and annotation by expert chemists is impractical for databases containing thousands (or more) data points.
Continue reading