Professor at University of Maryland Chemistry for AI and AI for Chemistry Co-founder, https://xitter.cf/t.co/PiGbhjKZcC

Joined June 2026
The wait ⏰ is over! Most structure prediction tools give you 1 answer. #RNA just doesn’t work that way - it's the ultimate shapeshifter. Today, we’re launching #RNAccess by Emergente Inc. with access for academic, nonprofit, and commercial researchers. RNAccess is powered by #RNAnneal, our physics-grounded #AI engine for #RNA structure prediction. By combining physics-based simulation with Generative AI, RNAnneal captures the structural flexibility that makes RNA both incredibly challenging—and incredibly powerful. The workflow is simple: 🧬Submit an RNA sequence of up to 100 nucleotides (long RNAs coming soon). ⚡Receive a thermodynamically ranked ensemble of high-accuracy 3D structures. 🔍Explore results through intuitive, interactive visualizations in your browser. 🦠Example: A key functional region of SARS-CoV-2 #RNA pictured below was predicted + visualized accurately in #RNAccess (with no prior knowledge, just physics!) We built #RNAccess to make serious RNA structure prediction more accurate, accessible, and useful for researchers working across all RNA work, from fundamental discovery to applied innovations. No local compute. No pipeline setup. No coding experience needed. Not even a GPU bill - we’ve got that covered 🙇 🎓Academic/non-commercial researchers: Receive a free batch of predictions every month, with pay-as-you-go options when you need more. 🏢Commercial teams: Start with a complimentary prediction, then talk with us about evaluation, confidential use, and larger-scale applications. 📍Try RNAccess: emergente-sci.com 📍Contact: [email protected] Come fold with us 💫 ------------------ It took a village to get this far. Thank you to @NIH-NIGMS, @NSF -Chemistry-CTMC, TEDCO, University of Maryland Institute for Health Computing, Institute for Physical Science and Technology, Montgomery County Government, UMD Chemistry and Biochemistry, @UMDscience, @UMmedschool and many others for financial and other support. Gratitude to leadership Bradley Maron, MD, Adam Porter, @VarshneyAmitabh, Mark T. Gladwin, MD, Martha Jurczak for their continued faith in our team at University of Maryland Institute for Health Computing and in Emergente Inc. - the first startup out of the IHC! And finally, huge thanks to our awesome scientific advisers Robert Copeland, Jonathan Dinman and John (Jay) Schneekloth for their guidance.
9
70
2
331
22,300
Big news! We at @UMDscience @UMmedschool are thrilled to be part of SEEDS, a new @ARPA_H supported consortium co-led by Georgetown University & Buck Institute to better understand & predict early protein changes associated with disease More about the science & team on Monday
Proteins are the tiny molecule machines that keep our cells working. When they misfold, cell function is disrupted which can lead to devastating disease. ARPA-H’s BIOGAMI teams are targeting that early window by decoding how proteins misfold and finding ways to steer them back on course. Earlier intervention could open new paths for Alzheimer’s, Parkinson’s, ALS, and more. Learn more about the teams pushing this work forward ➡️ arpa-h.gov/explore-funding/a…
5
202
Pratyush Tiwary retweeted
CTMD: Can molecules be accurately screened without waiting for free energy convergence? The most frustrating moment in virtual screening is often not ending up with an empty list, but rather getting hundreds of molecules that all look promising.
1
1
2
105
Presenting REDIAL: a zero-shot diagnostic to detect & quantify overparameterization in RNA language models. We find bigger RNA FMs are not automatically better—& propose path to more efficient, interpretable models. Preprint: biorxiv.org/content/10.64898… Code: github.com/tiwarylab/redial
1
3
19
1,309
4/5 The main finding is sobering: current RNA language models appear severely overparameterized relative to available RNA sequence diversity. In this domain, scale alone is not enough. Bigger is not automatically better.
1
2
207
5/5 The encouraging part: structure-guided pretraining improves the signal-to-noise ratio of learned base-pair couplings. We hope REDIAL helps guide RNA FMs that are not just larger, but more efficient, interpretable, and reliable for RNA therapeutics and de novo design.
2
177