🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

I really like this ProGen3 paper because, contrary to the title, I think it actually shows there is relatively little to be gained from massively scaling protein language models. 1/n

This is very irresponsible to give false promises like that to a millions of patients. If you are familiar with the complexity of the human diseases you know it's not possible. #unpopularopinion

Single-cell foundation models, trained on large-scale scRNA-seq datasets, are increasingly used for post-perturbation RNA-seq prediction. But how well do they actually perform? Our new paper from Turbine is now out in BMC Genomics. bmcgenomics.biomedcentral.com/articles/10.11… 1/5

Rudy Gilman Position, Padding and Predictions: A Deeper Look at Position Information in CNNs arxiv.org/abs/2101.12322

It was very surprising to found that during Ribonanza competition. One way to counter this was using masked convolution.

What a way to end #ICLR2025 in Singapore 🇸🇬 with an acceptance for PepTune in #ICML2025 in Vancouver 🇨🇦! 🥳 So many congratulations to my brilliant students Sophia Tang and Enol Z.! 🫶 A huge moment for our lab! 🙌 📜: arxiv.org/abs/2412.17780 A quick TLDR for PepTune, if

Delighted to share our latest work deciphering the landscape of chromatin accessibility and modeling the DNA sequence syntax rules underlying gene regulation during human development! biorxiv.org/content/10.110…. Read on for more 🧵 [1/16]

Really liked that reading

Synthetic biology could enable new types of programmable therapeutics. Our new preprint introduces synthetic protein circuits that selectively trigger cell death in Ras-mutant cancer cells, with interesting advantages compared to existing approaches. biorxiv.org/content/10.110…

I wrote a quick application note on Tomtom-lite, a Python implementation of the Tomtom algorithm for comparing PWMs against each other. This implementation can be 10-1000x faster and, as a Python function, can be integrated into your workflows easier. biorxiv.org/content/10.110…

Most biological processes, like stem cell differentiation, branch into multiple fates, but current trajectory inference methods only predict single paths.🌳To fix this, we introduce BranchSBM 🌿, from our unstoppable Sophia Tang ! 🌟 📜: arxiv.org/abs/2506.09007 💻:

Our paper is now accepted at Neural Networks! This work builds on our previous threads, updated with deeper analyses. We revisit brain-to-image reconstruction using NSD + diffusion models—and ask: do they really reconstruct what we perceive? Paper: doi.org/10.1016/j.neun… 🧵1/12

So you want to change transgene expression: just change your promoter, right? Changing the promoter increases RNA and thus protein levels. What more could be happening? [1/n] Well, promoters don’t just set RNA levels; they uniquely transform how RNAs are transmitted into protein

This evaluation of DNA design methods is very well written. If you're interested in the field, you should def take a look. Also, glad to see Ledidi performing so well! biorxiv.org/content/10.110…

Here's an idea. Generally in science, if something doesn't work, we say it doesn't work rather than claiming it does (most current virtual cell foundation models). Do the hard work & then make big claims. 2/

Anshul Kundaje (anshulkundaje@bluesky) The biggest meme is of course methods that don't compare to predicting basic baselines like 'mean perturbation effect' or a kmeans classifier, and are later found to be worse. Many others do compare to these and claim victory with some absolutely microscopy improvements over

Can an AI model predict perfectly and still have a terrible world model? What would that even mean? Our new ICML paper formalizes these questions One result tells the story: A transformer trained on 10M solar systems nails planetary orbits. But it botches gravitational laws 🧵

nature It's easy to target "sneaky researchers" instead of discussing the complex problem of sneaky reviewers. Shame.

I'm excited to share work on a research direction my team has been advancing: connecting machine learning derived genetic variant embeddings to downstream tasks in human genetics. This work was led by the amazing Divyanshi Srivastava! biorxiv.org/content/10.110…

People are always asking me if I know Adam