2026 may be the year AI starts to truly reason about biology.

AlphaFold helped close the sequence → structure gap.
The next frontier is sequence → functions.

Today, together with @genophoria and the team at @arcinstitute , we’re releasing BioReason-Pro — the first multimodal reasoning model for protein function prediction. Under the hood, BioReason-Pro is a multimodal reasoning LLM based on Qwen3-4B-Thinking. It takes protein embeddings, GO graph embeddings, and biological context, and then the language model autoregressively reasons in natural language from molecular evidence to function.

(1/5) I usually don't Twitter about preprints, but this one is an exception since I am extremely excited about it. We present DeepVelo ( biorxiv.org/content/10.110… ), a deep learning velocity estimation tool for scRNA-seq. (2/5) Existing techniques of inferring RNA velocity heavily depend on the strong assumptions of predefined RNA dynamics and constant kinetic rates across cells. However, many real-world datasets are complex and often contain multi-lineage cell systems with varying RNA dynamics.