What if a general LLM could become a biological instrument with the right fine-tuning? In @PNASNews, we present ADAR-GPT: a #GPT model trained step-by-step that learns ADAR targeting grammar, generalizes to new Alu sequences, and captures key constraints from the sequence alone.
#Codex for #HPC? We put it in the toughest arena: parallel code across heterogeneous hardware. Built a hardware-in-the-loop pipeline (compile→run→profile→fix) on real CPU+GPU toolchains to see what actually breaks. @OpenAIDevs
What if a general LLM could become a biological instrument with the right fine-tuning? In @PNASNews, we present ADAR-GPT: a #GPT model trained step-by-step that learns ADAR targeting grammar, generalizes to new Alu sequences, and captures key constraints from the sequence alone.
This is really exciting work that I am proud to be a part of. RNA is a remarkable material as it has structure and also carries information. My first paper published in 1969 modeled RNA structure in transferRNA.
The paper by @GalOren8 is about using AI to better predict messenger RNA editing sites so that the information it contains is more nuanced.
Biology is so full of amazing hacks.
🚨 New paper alert!
"Leveraging GPT Continual Fine-Tuning for Improved RNA Editing Site Prediction"
📍Now live at @MLGenX @ #ICLR2025
🧬🔥 We combine NLP + genomics to improve A-to-I RNA editing site detection using GPT models.
@MLGenX@MLevitt_NP2013 [7/]
🤖 More:
— No hand-crafted features
— Everything end-to-end from raw RNA + structure
— Ready for broader use in RNA therapeutics, gRNA design, etc.
@GalOren8 has done something remarkable in AI for Science by teaching ChatGPT the language of RNA. I am so proud to be involved in this project with him and Erez Levanon.
🎉 Excited to share our latest research published at AIDrugX @ NeurIPS 2024! We've developed an innovative approach to detect RNA editing sites using GPT fine-tuning. A thread 🧵👇 #NeurIPS2024#AI#Genomics#RNAEditing