PhyloPGM: boosting regulatory function prediction accuracy using evolutionary information

Abstract

Motivation The computational prediction of regulatory function associated with a genomic sequence is of utter importance in -omics study, which facilitates our understanding of the underlying mechanisms underpinning the vast gene regulatory network. Prominent examples in this area include the binding prediction of transcription factors in DNA regulatory regions, and predicting RNA–protein interaction in the context of post-transcriptional gene expression. However, existing computational methods have suffered from high false-positive rates and have seldom used any evolutionary information, despite the vast amount of available orthologous data across multitudes of extant and ancestral genomes, which readily present an opportunity to improve the accuracy of existing computational methods. Results In this study, we present a novel probabilistic approach called PhyloPGM that leverages previously trained TFBS or RNA–RBP binding predictors by aggregating their predictions from various orthologous regions, in order to boost the overall prediction accuracy on human sequences. Throughout our experiments, PhyloPGM has shown significant improvement over baselines such as the sequence-based RNA–RBP binding predictor RNATracker and the sequence-based TFBS predictor that is known as FactorNet. PhyloPGM is simple in principle, easy to implement and yet, yields impressive results. Availability and implementation The PhyloPGM package is available at https://github.com/BlanchetteLab/PhyloPGM Supplementary information Supplementary data are available at Bioinformatics online.

Graphylo: A deep learning approach for predicting regulatory DNA and RNA sites from whole-genome multiple alignments

“…We evaluated Graphylo against several state-of-the-art machine learning models that operate on single-species data: (1) FactorNet, 16 a top-performing single-species TFBS predictor in a recent DREAM competition 17 but limited to using only sequence information; (2) FactorNet+phastCons, a modified version of FactorNet that uses as input both the human sequence of interest and its interspecies conservation level (average PhastCons score); (3) RNATracker, 41 a tool initially designed for RNA subcellular localization analysis but also effective at TFBS prediction; and (4) PhyloPGM applied to FactorNet prediction values, 39 a probabilistic graphical model that combines predictions made on orthologous and ancestral sequences to improve a base FactorNet’s accuracy. Performance was measured using area under the receiver operating characteristic curve (AUROC) for the RBP binding task and the area under the precision-recall curve (AUPR) for the TF binding task on a held-out test set that were not used for training.…”

“…Comparing the performance of these models to Graphylo is a good measurement of the advantages Graphylo is taking from evolutionary information. We also compared Graphylo against PhyloPGM 39 which is a probabilistic graphical model that successfully incorporates predictions from different nodes in the phylogenetic tree.…”

“…For example, Alphafold2 38 uses multiple alignments of amino acid sequences to predict protein structure. For DNA sequences, there are few methods that used multiple sequence alignments including 33 , 39 but extracting features from the multiple sequence alignment within the context of evolution remains challenging.…”