Computational toxicology and ADMET, Cheminformatics, Applied AI/ML for drug design and discovery.
USA
Joined August 2009
- Tweets2.1K
- Following398
- Followers908
- Likes691
The paradox of living close to Philly: any event that's happening there requires at least an hour and half of travel time. While normal travel time to Philly is around 30 minutes... Because traffic...
ok, interesting work, but why one want to do a hierarchical (or any) clustering of 1B compounds? biorxiv.org/content/10.1101/…
The #Tox24 challenge is officially over. Several days of work resulted in 8 submissions and a 12th place overall, with a tie that almost put me into the top 10 (timing of submission does play a role in scoring). hdpharm.substack.com/p/tox24…
Some really nice points by @CynthiaRudin "Stop Explaining Black Box Machine Learning Models for High Stakes Decisions and Use Interpretable Models Instead". Some extracts below. TLDR: Explainable ML is not the same as interpretable. arxiv.org/abs/1811.10154
"Rather than producing explanations that are faithful to the original model, they show trends in how predictions are related to the features. Calling these ‘summaries of predictions’, ‘summary statistics’ or ‘trends’ rather than ‘explanations’ would be less misleading."[ibid]
"Interpretability/explainability is just as domain-specific as accuracy performance, so it is not clear why reviews of interpretability make any more sense than reviews of accuracy/performance. "[ibid]
"belief that there is always a trade-off between accuracy and interpretability has led many researchers to forgo the attempt to produce an interpretable model. This problem is compounded by the fact that researchers are now trained in deep learning, but not in interpretable ML"
Vladimir Chupakhin retweeted
So the race is really heating up to build a truly universal force field. This is one of those powerful ideas that people in the field of molecular simulation have been dreaming about for decades. What exactly is it and how far away are we? 1/n
A really good read (a reminder) about complexity of post-transnational modifications: onlinelibrary.wiley.com/doi/… + @uniprot actually has a extensive list for them ftp.uniprot.org/pub/database…
that's interesting about lysosomes : "APT1, which is mainly localized in the cytoplasm of yeast and mammalian cells, is a highly conserved α/β hydrolase containing the S-H-D catalytic triad and the G-X-S-X-G motif, with palmitoylated Ras proteins being its main substrates.602"
"Low palmitoylation of the mutant huntingtin (HTT) protein in the nervous system results in increased neurotoxicity and greater susceptibility to aggregate formation, which may induce Huntington's disease (HD).32"
Asn-Gly PTM not a surprise conceptually, but still an amazing reminder about complexity of biological processes.
