MOTIVATION: RNA secondary structure is often essential to function. Recent work has led to the development of high-throughput experimental probing methods for structure determination. Although structure is more conserved than primary sequence, much of the bioinformatics pipelines to connect RNA structure to function rely on nucleotide sequence alignments rather than structural similarity. There is a need to develop methods for secondary structure comparisons that are also fast and efficient to navigate the vast amounts of structural data. K-mer based similarity approaches are valued for their computational efficiency and have been applied for protein, DNA, and RNA primary sequences. However, these approaches have yet to be implemented for RNA secondary structure. RESULTS: Our method, bpRNA-CosMoS, fills this gap by using k-mers and length-weighted cosine similarity to compute similarity scores between RNA structures. bpRNA-CosMoS is built upon the bpRNA structure array, which represents the structural category of each nucleotide as a single-character structural code (e.g. hairpin=H, etc.). A structural comparison score is calculated through cosine similarity of the k-mer count vectors, generated from structure arrays. A major challenge with k-mer based methods is that they often ignore the length of the sequences being compared. We have overcome this with a length-weighted penalty that addresses cases of two RNAs of vastly different lengths. In addition, the use of "fuzzy counting" has added some optional flexibility to decrease the negative impact that small structural variations have on the similarity score. This results in a robust and efficient way to identify structural comparisons across large datasets. AVAILABILITY AND IMPLEMENTATION: The code and application guidelines of bpRNA-CosMoS are made available at github (https://github.com/BLasher113/bpRNA-CosMoS) and Zenodo (10.5281/zenodo.14715285).

Ribosomes are information-processing macromolecular machines that integrate complex sequence patterns in messenger RNA (mRNA) transcripts to synthesize proteins. Studies of the sequence features that distinguish mRNAs from long noncoding RNAs (lncRNAs) may yield insight into the information that directs and regulates translation. Computational methods for calculating protein-coding potential are important for distinguishing mRNAs from lncRNAs during genome annotation, but most machine learning methods for this task rely on previously known rules to define features. Sequence-to-sequence (seq2seq) models, particularly ones using transformer networks, have proven capable of learning complex grammatical relationships between words to perform natural language translation. Seeking to leverage these advancements in the biological domain, we present a seq2seq formulation for predicting protein-coding potential with deep neural networks and demonstrate that simultaneously learning translation from RNA to protein improves classification performance relative to a classification-only training objective. Inspired by classical signal processing methods for gene discovery and Fourier-based image-processing neural networks, we introduce LocalFilterNet (LFNet). LFNet is a network architecture with an inductive bias for modeling the three-nucleotide periodicity apparent in coding sequences. We incorporate LFNet within an encoder-decoder framework to test whether the translation task improves the classification of transcripts and the interpretation of their sequence features. We use the resulting model to compute nucleotide-resolution importance scores, revealing sequence patterns that could assist the cellular machinery in distinguishing mRNAs and lncRNAs. Finally, we develop a novel approach for estimating mutation effects from Integrated Gradients, a backpropagation-based feature attribution, and characterize the difficulty of efficient approximations in this setting.

The analysis of RNA structure is an important problem because RNA plays essential roles throughout all living things. Increasingly, RNA plays pivotal roles in biotechnology including RNA silencing, guide RNAs for CRISPR, ligand-binding aptamers, and drug design. While several studies have used deep learning models for analyzing RNA structure, these approaches rely on the outputs of traditional dynamic programming algorithms based on known thermodynamic energy parameters during training. The challenge remains to train neural networks to independently learn thermodynamic principles, so that potentially new biology can be uncovered. Here we present DeepSloop, a deep recurrent network capable of distinguishing hairpin-forming sequences from non-hairpin-forming, and that is trained purely on class-labeled nucleotide sequences. We demonstrate that DeepSloop can accurately classify hairpin stem-loops, and thus learned a distinct internal representation of structure that includes destabilizing energies of bulges and internal loops.

MOTIVATION: Predicting the secondary structure of an ribonucleic acid (RNA) sequence is useful in many applications. Existing algorithms [based on dynamic programming] suffer from a major limitation: their runtimes scale cubically with the RNA length, and this slowness limits their use in genome-wide applications. RESULTS: We present a novel alternative O(n3)-time dynamic programming algorithm for RNA folding that is amenable to heuristics that make it run in O(n) time and O(n) space, while producing a high-quality approximation to the optimal solution. Inspired by incremental parsing for context-free grammars in computational linguistics, our alternative dynamic programming algorithm scans the sequence in a left-to-right (5'-to-3') direction rather than in a bottom-up fashion, which allows us to employ the effective beam pruning heuristic. Our work, though inexact, is the first RNA folding algorithm to achieve linear runtime (and linear space) without imposing constraints on the output structure. Surprisingly, our approximate search results in even higher overall accuracy on a diverse database of sequences with known structures. More interestingly, it leads to significantly more accurate predictions on the longest sequence families in that database (16S and 23S Ribosomal RNAs), as well as improved accuracies for long-range base pairs (500+ nucleotides apart), both of which are well known to be challenging for the current models. AVAILABILITY AND IMPLEMENTATION: Our source code is available at https://github.com/LinearFold/LinearFold, and our webserver is at http://linearfold.org (sequence limit: 100 000nt). SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

MicroRNAs are conserved, endogenous small RNAs with critical post-transcriptional regulatory functions throughout eukaryota, including prominent roles in development and disease. Despite much effort, microRNA annotations still contain errors and are incomplete due especially to challenges related to identifying valid miRs that have small numbers of reads, to properly locating hairpin precursors and to balancing precision and recall. Here, we present miRWoods, which solves these challenges using a duplex-focused precursor detection method and stacked random forests with specialized layers to detect mature and precursor microRNAs, and has been tuned to optimize the harmonic mean of precision and recall. We trained and tuned our discovery pipeline on data sets from the well-annotated human genome, and evaluated its performance on data from mouse. Compared to existing approaches, miRWoods better identifies precursor spans, and can balance sensitivity and specificity for an overall greater prediction accuracy, recalling an average of 10% more annotated microRNAs, and correctly predicts substantially more microRNAs with only one read. We apply this method to the under-annotated genomes of Felis catus (domestic cat) and Bos taurus (cow). We identified hundreds of novel microRNAs in small RNA sequencing data sets from muscle and skin from cat, from 10 tissues from cow and also from human and mouse cells. Our novel predictions include a microRNA in an intron of tyrosine kinase 2 (TYK2) that is present in both cat and cow, as well as a family of mirtrons with two instances in the human genome. Our predictions support a more expanded miR-2284 family in the bovine genome, a larger mir-548 family in the human genome, and a larger let-7 family in the feline genome.