Anh-Tuan PhamI had an incredible time at #ACSSpring2025! The conference was filled with groundbreaking research and cutting-edge advancements in chemistry and biology, and I was truly amazed by the depth of innovation on display.
Egon Willighagenlast speaker in this session is Tomoyuki Miyao talking about "Three-dimensional molecular representation in machine learning models"
Artem Cherkasov is now talking, and that "big bang" figure I mentioned, is actually from this paper (AC is last author of the 6D paper): https://doi.org/10.1038/s41589-022-01233-x
He talks us through deep docking. to show disruptions. I wonder... will they name the @cdk in this talk as open source cheminformatics toolkit, like in the paper?
PB talks about large data sets and show Butina clustering. I had to look that up, and since from 1999, that should have showed up on my radar at some point, but don't remember is. https://doi.org/10.1021/CI9803381
@wdscholia provides context to this paper: https://scholia.toolforge.org/work/Q56337260
we've just had a break and now speaking is Pedro Ballester, taking about benchmarking, and had a nice slide showing various data sets in a 2d data space, a bit similar to that in the recent "big bang" Figure 3 from https://doi.org/10.1016/j.drudis.2025.104341 but then with datasets instead of molecules
AL shows the used of SHAP for global feature importance, showing which parts of the molecule are important to the model predictivity
I just finished my own talk in this session (slides will be online later), and now talking is Alec Lamens.
They are now talking about explainability and interpretability in AI (versus black box AI)
OE invites everything to play with REINVENT, and mentions https://github.com/MolecularAI
REINVENT has seen several publications; he shows titles of some 8 or 9 articles
OE about Open Science in pharma: reproducibility, transparency, collaboration, open source, open data, open benchmarks
"you can understand code progress" "jungle of methods out there"
Shout out by Ola to RDKit and CDK, PubChem, GROMACS, PyTorch, and SkikitLearn
OE discusses the drug discovery process: design, make, test, and analyze (DMTA) and how we can accelerate this, an interative process. We now make the transition to seeing how AI can help here