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...
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
"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]
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
"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]
"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]
"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"
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
"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"
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
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"