Top Tweets for #RNAGenesis

over 1 year ago

Our collaborative work, #RNAGenesis, led by @ZaixiZhang and team, delivers a Generalist RNA Foundation Model (full weights released!), end-to-end pipeline that understand, design, validate functional RNAs—from nM affinity aptamers to CRISPR scaffolds that lift editing efficiency up to 2.5×, when tested across nuclease, base-editing, and prime-editing. Let's move beyond prediction tasks, start applying generative AI design for RNA therapeutics! Please check out our preprint: https://t.co/LXfhA263X8 Full model released at: https://t.co/rrsbn2GWLa Thanks to all of amazing collaborators especially at #Princeton and #Stanford #RNA #CRISPR #GeneEditing #RNAtherapeutics #DrugDiscovery

Zaixi Zhang

@ZaixiZhang

over 1 year ago

🚀 Introducing 𝗥𝗡𝗔𝗚𝗲𝗻𝗲𝘀𝗶𝘀 — a 𝗚𝗲𝗻𝗲𝗿𝗮𝗹𝗶𝘀𝘁 𝗙𝗼𝘂𝗻𝗱𝗮𝘁𝗶𝗼𝗻 𝗺𝗼𝗱𝗲𝗹 for 𝗳𝘂𝗻𝗰𝘁𝗶𝗼𝗻𝗮𝗹 𝗥𝗡𝗔 𝘁𝗵𝗲𝗿𝗮𝗽𝗲𝘂𝘁𝗶𝗰𝘀, unifying sequence understanding, de novo design, and 3D structure prediction. From 𝗔𝗽𝘁𝗮𝗺𝗲𝗿𝘀 to 𝗚𝗲𝗻𝗲 𝗘𝗱𝗶𝘁𝗶𝗻𝗴, RNAGenesis powers next-gen RNA engineering across modalities. 🧬 Key results: ➡ De novo aptamer design with 𝗻𝗮𝗻𝗼𝗺𝗼𝗹𝗮𝗿 𝗞𝗗 (as low as 4.02 nM) ➡ sgRNA scaffolds boosting 𝗴𝗲𝗻𝗲 𝗲𝗱𝗶𝘁𝗶𝗻𝗴 𝗲𝗳𝗳𝗶𝗰𝗶𝗲𝗻𝗰𝘆 𝘂𝗽 𝘁𝗼 𝟮.𝟱× (CRISPR-Cas9, Base Editing, Prime Editing) ➡ SOTA on 𝟭𝟭 𝗼𝗳 𝟭𝟯 tasks in the BEACON benchmark ➡ Outperforms 𝗔𝗹𝗽𝗵𝗮𝗙𝗼𝗹𝗱𝟯 (structure prediction), 𝗥𝗵𝗼𝗗𝗲𝘀𝗶𝗴𝗻 (inverse folding), and 𝗥𝗡𝗔-𝗙𝗿𝗮𝗺𝗲𝗙𝗹𝗼𝘄 (de novo structure generation) ➡ Built 𝗥𝗡𝗔𝗧𝘅-𝗕𝗲𝗻𝗰𝗵 — a focused benchmark for RNA therapeutics with 100K+ validated sequences; RNAGenesis outperforms 𝗘𝘃𝗼𝟮 and 𝗥𝗡𝗔-𝗙𝗠 across ASO, siRNA, shRNA, circRNA, and UTR tasks 🧠 Powered by: A hybrid 1B-param model: BERT-style encoder + latent diffusion decoder Inference-time alignment (gradient guidance + beam search) Multi-modal design unifying sequence, structure, and function 🧪 Wet-lab validation confirms high-affinity aptamer binding and improved gene-editing activity 🧮 Computational analysis reveals stronger G–C pairing, more hydrogen bonds, and lower MFE/MMBPSA in designed RNAs 📌 Explore: 🧾 Paper: https://t.co/XLMGpJvHJL 💻 Code: https://t.co/N3c2PUfPeD 🧪 Benchmark: RNATx-Bench Team effort from @PrincetonAInews @StanfordMed @ZJU_China @PKU1898 @lecong @MengdiWang10 and more. #RNA #CRISPR #GeneEditing #RNAtherapeutics #DrugDiscove

ZaixiZhang's tweet photo. 🚀 Introducing 𝗥𝗡𝗔𝗚𝗲𝗻𝗲𝘀𝗶𝘀 — a 𝗚𝗲𝗻𝗲𝗿𝗮𝗹𝗶𝘀𝘁 𝗙𝗼𝘂𝗻𝗱𝗮𝘁𝗶𝗼𝗻 𝗺𝗼𝗱𝗲𝗹 for 𝗳𝘂𝗻𝗰𝘁𝗶𝗼𝗻𝗮𝗹 𝗥𝗡𝗔 𝘁𝗵𝗲𝗿𝗮𝗽𝗲𝘂𝘁𝗶𝗰𝘀, unifying sequence understanding, de novo design, and 3D structure prediction.
From 𝗔𝗽𝘁𝗮𝗺𝗲𝗿𝘀 to 𝗚𝗲𝗻𝗲 𝗘𝗱𝗶𝘁𝗶𝗻𝗴, RNAGenesis powers next-gen RNA engineering across modalities.
🧬 Key results:
➡ De novo aptamer design with 𝗻𝗮𝗻𝗼𝗺𝗼𝗹𝗮𝗿 𝗞𝗗 (as low as 4.02 nM)
➡ sgRNA scaffolds boosting 𝗴𝗲𝗻𝗲 𝗲𝗱𝗶𝘁𝗶𝗻𝗴 𝗲𝗳𝗳𝗶𝗰𝗶𝗲𝗻𝗰𝘆 𝘂𝗽 𝘁𝗼 𝟮.𝟱× (CRISPR-Cas9, Base Editing, Prime Editing)
➡ SOTA on 𝟭𝟭 𝗼𝗳 𝟭𝟯 tasks in the BEACON benchmark
➡ Outperforms 𝗔𝗹𝗽𝗵𝗮𝗙𝗼𝗹𝗱𝟯 (structure prediction), 𝗥𝗵𝗼𝗗𝗲𝘀𝗶𝗴𝗻 (inverse folding), and 𝗥𝗡𝗔-𝗙𝗿𝗮𝗺𝗲𝗙𝗹𝗼𝘄 (de novo structure generation)
➡ Built 𝗥𝗡𝗔𝗧𝘅-𝗕𝗲𝗻𝗰𝗵 — a focused benchmark for RNA therapeutics with 100K+ validated sequences; RNAGenesis outperforms 𝗘𝘃𝗼𝟮 and 𝗥𝗡𝗔-𝗙𝗠 across ASO, siRNA, shRNA, circRNA, and UTR tasks
🧠 Powered by:
A hybrid 1B-param model: BERT-style encoder + latent diffusion decoder
Inference-time alignment (gradient guidance + beam search)
Multi-modal design unifying sequence, structure, and function
🧪 Wet-lab validation confirms high-affinity aptamer binding and improved gene-editing activity
🧮 Computational analysis reveals stronger G–C pairing, more hydrogen bonds, and lower MFE/MMBPSA in designed RNAs
📌 Explore:
🧾 Paper: https://t.co/XLMGpJvHJL
💻 Code: https://t.co/N3c2PUfPeD
🧪 Benchmark: RNATx-Bench
Team effort from @PrincetonAInews @StanfordMed @ZJU_China @PKU1898 @lecong @MengdiWang10 and more.
#RNA #CRISPR #GeneEditing #RNAtherapeutics #DrugDiscove

17K

Top Tweets for #RNAGenesis

Last Seen Hashtags on Sotwe

Trends for you

Most Popular Users