New machine learning method improves computational prediction of RNA structures

The researchers present ARES (Atomic Rotationally Equivariant Scorer) – a machine learning method that significantly improves the computational prediction of RNA structures over previous approaches. Like proteins, RNA molecules twist and fold into complex three-dimensional shapes that are crucial for their function.

Understanding these structures could help uncover the biological functions of RNA, including non-coding RNA, and pave the way for the discovery of new drugs for diseases that remain incurable. However, the experimental resolution of RNA structure remains a challenge despite decades of efforts, and only a few RNA structures are currently known.

Additionally, using machine learning to predict RNA structure has proven to be much more difficult – and less successful – than for predicting protein structure. To address these challenges, Raphael Townshend and his colleagues developed ARES, a deep neural network capable of consistently producing precise RNA structural models, although they were trained using data for only 18 structures. recently experimentally determined RNA.

According to the authors, ARES has clearly outperformed other computational approaches in the challenge of predicting the structure of RNA-Puzzles at the community level. Townshend et al. note that ARES’s performance is particularly noteworthy as it has learned to make its predictions based only on atomic structure and does not make prior assumptions about RNA-specific structural characteristics that might be important, such as as base pairs, nucleotides or hydrogen bonds. In addition, it was able to accurately predict larger and more complex RNA structures than those on which it was trained.

ARES is still below the level compatible with atomic resolution or sufficient to guide identification of key functional sites or drug discovery efforts, but Townshend et al. have made notable progress in an area that has proven reluctant to transformative breakthroughs. “

Kevin weeks