ePOSSUM

ePOSSUM evaluates the impact of DNA variants on transcription factor (TF) binding sites to assess potential effects on gene regulation.

Key Features:

• In Silico Modeling: Employs position-specific scoring matrices (PSSMs) derived from JASPAR, HT-SELEX-generated models, and protein binding microarrays (PBMs) to model variant effects on TF binding.
• Systematic Comparison: Systematically compares PSSM-based models against experimentally verified in vivo TF binding sites from ENCODE ChIP-seq data.
• Performance Assessment: Assesses model performance using receiver operating characteristic (ROC) analysis and reports area-under-curve (AUC) scores, noting that only a subset reach an AUC of 0.7 or higher.
• Bayes Classifier: Incorporates a Bayes classifier to provide probabilistic evaluation of how genetic alterations affect TF binding within user-defined sequences.
• Reliability Information: Evaluates prediction reliability by referencing a test set of experimentally confirmed binding sites.

Scientific Applications:

• Gene regulation analysis: Characterizing mechanisms of gene regulation by quantifying variant-induced changes in TF binding.
• Disease variant interpretation: Prioritizing and interpreting regulatory DNA variants that may contribute to disease through altered TF binding.
• Model benchmarking and refinement: Comparing and refining TF binding models by benchmarking JASPAR, HT-SELEX, and PBM-derived PSSMs against ENCODE ChIP-seq.

Methodology:

Uses PSSMs from JASPAR, HT-SELEX, and PBM-derived models; compares them to ENCODE ChIP-seq TF binding sites; performs ROC analysis to compute AUC; and applies a Bayes classifier to assess variant effects.