RNA-FrameFlow for de novo 3D RNA Backbone Design

Rishabh Anand, Chaitanya K. Joshi, Alex Morehead, Arian Rokkum Jamasb, Charles Harris, Simon V Mathis, Kieran Didi, Bryan Hooi, Pietro Lio

Keywords: RNA Structure, RNA Design, Generative Modelling, Flow Matching, Geometric Deep Learning

TL;DR: 3D generative model for RNA backbone structures

Abstract: We introduce RNA-FrameFlow, the first generative model for 3D RNA backbone design. We build upon $SE(3)$ flow matching for protein backbone generation and focus on establishing RNA-specific data augmentations and evaluation protocols. Our formulation of rigid-body frames and loss functions account for larger, more conformationally flexible RNA backbones (13 atoms) vs. proteins (4 atoms). Towards tackling the lack of diversity in 3D RNA datasets, we explore training with structural clustering and cropping augmentations. Additionally, we define a suite of in-silico evaluation metrics to measure whether designed RNAs are globally self-consistent (via inverse folding followed by forward folding) and locally recover RNA-specific structural descriptors. The most performant version of RNA-FrameFlow generates locally realistic backbone structures of 40-150 nucleotides that are 41% globally self-consistent on average (scTM $\geq$ 0.45), with fast sampling speeds of $\sim$4 seconds per backbone.

Submission Number: 126