Vladimir Chupakhin

Computational toxicology and ADMET, Cheminformatics, Applied AI/ML for drug design and discovery.

[Vladimir Chupakhin @chupvl]

@jdudley @AnthropicAI :) whoa, you hit those just now

[Vladimir Chupakhin @chupvl]

HDPharm: Monthly Insights, January 2025 https://t.co/FGw1wF0SHp

[Vladimir Chupakhin @chupvl]

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...

[Vladimir Chupakhin @chupvl]

ok, interesting work, but why one want to do a hierarchical (or any) clustering of 1B compounds? https://t.co/JtMpqQ36hS

[Vladimir Chupakhin @chupvl]

@egor__marin @LindorffLarsen Looks at those MD archives :)

[Vladimir Chupakhin @chupvl]

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). https://t.co/Lqm7EGF11n

[Vladimir Chupakhin @chupvl]

@TimothyDuignan Yeah, it's not possible from first principles. And IMO why on earth do you need to simulate a neuron using first principles?

[Vladimir Chupakhin @chupvl]

"recent 2018 FICO Explainable ML Challenge exemplified the blind belief in the myth of the accuracy/interpretability tradeoff for a specific domain, namely credit scoring. However, there was no performance difference between interpretable models and explainable models" [ibid]

[Vladimir Chupakhin @chupvl]

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. https://t.co/RiT7wF1pbh

[Vladimir Chupakhin @chupvl]

"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]

[Vladimir Chupakhin @chupvl]

"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]

[Vladimir Chupakhin @chupvl]

"Saliency maps can be useful to determine what part of the image is being omitted by the classifier, but this leaves out all information about how relevant information is being used" [ibid]

[Vladimir Chupakhin @chupvl]

"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"

[Tim Duignan @TimothyDuignan]

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

[Vladimir Chupakhin @chupvl]

@KRHornberger Hey-hey, you forgot GSK3B, MTOR

[Vladimir Chupakhin @chupvl]

"O-GlcNAcylation of α-synuclein affects its phosphorylation and blocks the toxicity of α-synuclein, suggesting that an increase in O-GlcNAcylation may prevent α-synuclein aggregation (Figure 19).1182"

[Vladimir Chupakhin @chupvl]

A really good read (a reminder) about complexity of post-transnational modifications: https://t.co/tQyGS3egAO + @uniprot actually has a extensive list for them https://t.co/tkjeKf1zU0

[Vladimir Chupakhin @chupvl]

"Interestingly, the level of O-GlcNAcylation is reduced in AD patients.1178"

[Vladimir Chupakhin @chupvl]

"Genetic defects in glycosylation are often embryonic lethal.1136"

[Vladimir Chupakhin @chupvl]

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"