RAFA: RNA all-atom structure reconstruction from sparse anchor coordinates

Abstract

SUMMARY: We present RAFA, a C ++ command-line tool for RNA all-atom structure reconstruction from sparse or partial coordinates. RAFA targets settings in which coarse-grained modelling, low-resolution fitting, or partial experimental interpretation provides RNA anchor atoms but not a complete atomic model. The method retrieves experimentally observed 3-5 nt all-atom RNA fragments, fits them to local anchors, combines overlapping fragment-derived coordinate estimates by weighted consensus, and routes sparse and rich inputs through distinct reconstruction paths. In an independent RNA3DB train/test evaluation, no library-test pair exceeded 80% global sequence identity. RAFA achieved lower all-atom root-mean-square deviation (RMSD) values than Arena in 73.3% of target-mode comparisons, with the largest gains in input modes with the fewest structural constraints. AVAILABILITY AND IMPLEMENTATION: RAFA is available at https://github.com/wangleiofficial/RAFA and archived at https://doi.org/10.5281/zenodo.20951205. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.