🧬 😴 TIRED: Scaling protein models to billions of parameters hoping they'll memorize all of evolution and generalize beyond.🔥 WIRED: Smart retrieval-augmented models that dynamically access what they need from sequence databases.

RT @schwabpa: Save the date! Machine Learning for Drug Discovery (MLDD) is happening soon on Monday 30 June, 2025. MLDD aims to bring toge….

RT @AvivSpinner: I wonder if the gym membership for ProteinGym covers RNAGym as well?? oh wait, it's all free and 100% open source!.

Links:.🔗 Paper: 💻 Code: 9/9.

Congratulations to the entire RNAGym team! @rohitarorayyc @mvrfalo @christian_choe_ @c_sheare @aaronkollasch @fiona__qu @ruben_weitzman @artemg2718 @sarahgurev Erik Xie @deboramarks .8/9.

The moderate performance across all tasks reveals exciting opportunities! Key directions: RNA-specific training data, integrating structure-function relationships, and improving non-canonical base pair prediction. RNAGym provides a standardized foundation for progress. 7/9.

🌀 Tertiary structure: 215 diverse 3D structures from recent PDB entries. NuFold leads monomers (0.393 TM-score), AlphaFold3 dominates complexes (0.381 TM-score). Non-Watson-Crick interactions remain a major challenge for all methods.6/9

🔗 Secondary structure: 901k chemical mapping profiles using DMS & 2A3 reactivity. EternaFold achieves top performance (0.656 F1-score), closely followed by CONTRAfold & Vienna. Traditional thermodynamic methods are still competitive with newer deep learning approaches.5/9

🔬 Fitness prediction: 70 assays across tRNA, ribozymes, aptamers & mRNAs (1M+ mutations). Evo 2 performs best overall, but performance varies dramatically by RNA type: RNA-FM excels at tRNA/aptamers while Evo 2 leads mRNA tasks. Lots of room for improvement across the board!.4/9

RNAGym tackles three essential RNA prediction tasks: 🔬 Fitness prediction: How mutations affect RNA function 🔗 Secondary structure: Base-pairing patterns 🌀 Tertiary structure: 3D molecular architecture.All evaluated zero-shot to test true generalization!.3/9.

Why do we need this? RNA modeling faces major challenges: limited experimental data (<1% of PDB entries), inherently less stable structures than proteins, and evaluation has been scattered across different studies with varying approaches. 2/9.

🚨 New paper 🚨 RNA modeling just got its own Gym! 🏋️ Introducing RNAGym, large-scale benchmarks for RNA fitness and structure prediction. 🧵 1/9

RT @AvivSpinner: What would our data landscapes look like if we could biochemically characterize 100s, 1000s, 10^n evolutionary sequences?….

RT @lood_ml: As sequence databases get bigger and more diverse, retrieval-based methods provide an interesting alternative to scaling succe….

RT @KevinKaichuang: End-to-end differentiable homology search for protein fitness prediction. @ruben_weitzman @lood_ml @yaringal @debora….

More details in @RubenWeitzman's excellent tweetorial above and in a new blog post: paper:

Introducing Protriever — SOTA fitness prediction for sequence-based models. Homolog retrieval orders of magnitude faster than prior approaches. These architectures will be essential building blocks for next-gen protein models!.

Future-proofed vaccine design with generative models that predict viral evolution 👇.

RT @nooryoussef03: 🚨 New in @ImmunityCP !.EVE-Vax, an AI model that anticipates future viral evolution and designs antigens to proactively….

Even simple methods leveraging these 2 modalities significantly outperform billion-parameter sequence-only models. So, what's next? Better retrieval, advanced multimodal approaches, & alignment. Read more: #BioTech #AI #pLMs.