Computational biologist; Asst. Prof. at University of Wisconsin-Madison; Investigator at Morgridge Institute
Jun 25 • 10 tweets • 3 min read
I'm sharing some first impressions of the ESM3 paper in this thread. The model and generative programming results look great, and I may come back to those later. 1/
I dislike the framing "Simulating 500 million years of evolution". That is confusing from a science communications perspective. ESM3 does not literally simulate evolution. Scientists should be precise when talking about AI simulating evolution, which sounds scary. 2/
Oct 21, 2021 • 9 tweets • 4 min read
Our preprint "Evaluating scalable supervised learning for synthesize-on-demand chemical libraries" is now live on ChemRxiv doi.org/10.33774/chemr…
Work with @SpencerEricksen @Shengchao_Liu and other great @UWMadison collaborators. 1/n
We were motivated by a problem that @StevenKearnes excellently describes in arxiv.org/abs/2009.00707: retrospective testing of ML models for chemical bioactivity against a target of interest can't tell us how they would perform in a real prospective setting. 2/n
Feb 1, 2019 • 9 tweets • 4 min read
Our preprint on gene regulatory network inference from single-cell RNA-seq is up doi.org/10.1101/534834
The SCINGE method uses ensembles of Granger Causality models to estimate regulator-gene interactions from single-cell data with pseudotimes. @atulpdeshpande@ron_stewart2 1/9
Granger Causality works well for long time series, which we rarely have with bulk gene expression data. Single-cell data is appealing because we have many samples making a long (pseudo) time series. However, the samples are irregularly-spaced in pseudotime. 2/9
