PhyloPGM

PhyloPGM leverages evolutionary information to improve regulatory function prediction of genomic sequences by probabilistically aggregating outputs from sequence-based predictors across orthologous and ancestral genomes.

Key Features:

• Probabilistic aggregation framework: Systematically combines outputs from multiple sequence-based predictors using a probabilistic model.
• Use of orthologous and ancestral genomic data: Incorporates predictions from extant and ancestral orthologous regions to inform human-sequence predictions.
• Integration of pre-trained predictors: Aggregates outputs from previously trained predictors for transcription factor binding sites (TFBS) and RNA–binding protein (RBP) interactions.
• Application to DNA and RNA regulatory signals: Targets both TFBS prediction within DNA regulatory regions and RNA–protein interaction prediction.
• Improved predictive performance: Demonstrated substantial improvements over established baselines such as RNATracker and FactorNet.
• Reduced false-positive rates: Utilizes evolutionary conservation across orthologs to refine predictions and lower false positives relative to traditional methods.

Scientific Applications:

• TFBS prediction: Enhances identification of transcription factor binding sites within DNA regulatory regions using orthologous evidence.
• RNA–RBP interaction prediction: Improves prediction of RNA–binding protein binding sites on RNA sequences by aggregating cross-species predictions.
• Regulatory genomics: Supports studies of transcriptional and post-transcriptional gene regulation by integrating evolutionary signals.
• -Omics analyses: Increases reliability of regulatory function predictions in genome-wide and transcriptome-wide investigations.

Methodology:

Aggregate predictions from previously trained sequence-based predictors across orthologous and ancestral genomic regions using a probabilistic framework, and evaluate performance by comparison to baselines such as RNATracker and FactorNet.