RNA-Ligand Prediction Model

Overflow from /projects since that section got a bit unwieldy

This project traces back to spring 2025, when I began replicating existing RNA-ligand prediction models as an exercise to learn ML in a field especially resonant to me (RNA drugs could treat EDS and the prediction models reminisced of AlphaFold).

It didn’t become a research project, though, until mid-summer. On a whim, I borrowed a new feature from the protein literature not used in RNA, and the model actually improved! I became wholly obsessed with the idea that there might be room to really push the state of the art, and after many months of adapting insights from near and far fields (and an embarassing amount of student Colab credits), that pipe dream really came true. :)

• How to discover RNA drugs with AI (brief explainer)
• Enabling RNA-drug discovery with RiboLead (paper draft)
• You don’t need permission to make progress (early reflections + the process)
• Key model architecture / train scripts on GitHub
• Web interface for drug screening (in demo stage, integrating model)
• Refactored source code with portable, one-shot colab setup scripts