Physics-informed deep learning has been getting quite some attention in physics (besides ML), see eg. nature.com/articles/s4225… - what are (or could be) examples of such constraints and models in cell biology? Thinking population models, RNA velocity, signaling pathways etc.

Weekend project: LLM-powered QA for cooking-related topics with reference recipes. Thanks Andrew White 🐦‍⬛ for the inspiration!

The order of features in a neural net doesn’t affect its function. But, hypernetworks & models that process neural net weights don’t account for this! Turns out we can encode this permutation invariance in neural functional networks.

Tweetorial time! We Recursion mapped consequences of #CRISPR screening of >17K human genes, found a systematic bias confounding all CRISPR screens, traced its molecular cause, and propose a debiasing algorithm.

My team at #NVIDIA is hiring for multiple life sciences applied deep learning scientists. Help us build #BioNeMo! - Protein Structure & Design link.mlgill.co/nvidia-protein… - Cheminformatics & Virtual Screening link.mlgill.co/nvidia-cheminf… - Geometric Deep Learning link.mlgill.co/nvidia-geometr…

A reminder that Elsevier made $10.5 BILLION in 2022 from selling your academic journals and articles behind paywalls, and make more profit than Amazon, Google, and Apple every year… And paid the academics who wrote the articles $0 And paid the reviewers of the articles $0

🤌🏼🤌🏼

Nature Methods Lucas Pelkmans Happy that CAMPA is out Nature Methods - led by Hannah Spitzer & Scott Berry w Lucas Pelkmans we developed a conditional autoencoder for multiplexed image analysis, learning subcellular landmarks consistent across perturbations. github.com/theislab/campa nature.com/articles/s4159…

Coming to American Society for Cell Biology #ascb2023 this December? Consider submitting a talk to the subgroup on Deep Learning and Artificial Intelligence in Cell Imaging! Since it's on my home turf of Boston, I can promise a wicked awesome time. Details in link! ascb.org/cellbio2023/pr…

Delighted to share our latest preprint on Image2Omics - an approach to connect cell shape and multi-omics! Huge shoutout to lead authors Rahil Mehrizi and Arash Mehrjou! Preprint: arxiv.org/abs/2306.09391 Code: github.com/GSK-AI/image2o… Visit gsk.ai to learn more!

Excited to introduce scKINETICS, an algorithm for joint inference of gene regulatory mechanism and cellular velocity, co-developed by the amazing duo of Cassandra Burdziak and Julia Zhao academic.oup.com/bioinformatics…

As part of our meetup at #ICML2023 on Wednesday, we’ll be hosting a panel with Cuong Nguyen (@gsk), Alan Cheng (@merck), @imranshaque (@recursionpharma), and Charlotte Bunne (@eth), moderated by Dan RSVP to secure your spot below: partiful.com/e/RQovXQfUfOAx…

Last week at the #ICML2023 community meetup, we hosted a panel on the topic of benchmarks and their importance in fostering the development of impactful AI models in drug discovery. The recording is now available here: youtube.com/watch?v=g2rHFQ…

Pretty amazing that even Sutton & Barto, Bertsekas, and David Silver with their high quality teaching materials could not achieve what a single OpenAI rumor did: make people on Twitter learn and talk about Q-learning!

Are you attending #NeurIPS2023?💡 I am actively recruiting for a number of open roles at different levels in my team at GSK.ai If you are interested in leveraging AI to understand and advance solutions in health 🏥 please reach out! ✉️

🎉🪩🎄 Check out our new PerturbView method for easy optical perturbation screens in primary cells and tissues. Takamasa Kudo has written a great summary in the quoted thread. Check it out! biorxiv.org/content/10.110…