LinearAlifold: Linear-Time Consensus Structure Prediction for RNA Alignments

Predicting the consensus structure of a set of aligned RNA homologs is a convenient method to find conserved structures in an RNA genome, which has applications in SARS-CoV-2 diagnostics and therapeutics. However, the state-of-the-art algorithm for this task, RNAalifold, is prohibitively slow for long sequences, due to a cubic scaling with the sequence length, and even slower when analyzing many such sequences, due to a superlinear scaling with the number of homologs, taking 4 days on 200 SARS-CoV variants. We present LinearAlifold, an efficient algorithm for folding aligned RNA homologs that scales linearly with both the sequence length and the number of sequences, based on our recent work LinearFold that folds a single RNA in linear time. Our work is orders of magnitude faster than RNAalifold (e.g., 0.5 hours on the above 200 sequences or 316 times speedup) and achieves comparable accuracies compared to a database of known structures. More interestingly, LinearAlifold's prediction on SARS-CoV-2 correlates well with experimentally determined structures, outperforming RNAalifold. Finally, LinearAlifold supports three modes: minimum free energy (MFE), partition function, and stochastic sampling, each of which takes under an hour for hundreds of SARS-CoV variants, while only the MFE mode of RNAalifold works for them, taking days or weeks.