Updated binding model for a splicing factor, PTBP1 (Polypyrimidine Tract Binding Protein1)

RNA-binding proteins (RBPs) are central regulators of post-transcriptional gene expression, and identifying their RNA targets is essential for understanding transcriptome dynamics. Predictive computational models enable the identification of RBP binding sites from RNA sequences in silico , complementing experimental assays. We previously developed a PTBP1 binding model based on a Hidden Markov Model (HMM) that utilized overlapping nucleotide triplets but was limited due to the assumption of independence between triplets. In this study, we present an improved PTBP1 binding model using Transcription Factor Flexible Models (TFFMs), which relax this assumption and accommodate flexible-length motifs. Trained on PTBP1 CLIP-seq data from HeLa cells, the TFFM-based approach captures positional dependencies among nucleotides, consistent with known PTBP1 binding motifs, and achieves improved predictive performance. The updated model shows a stronger correlation with experimental binding affinities (r = −0.95) compared to the previous HMM (r = −0.91) and maintains specificity by correctly scoring non-binding RNA substrates. This work highlights the advantage of flexible, context-aware modeling frameworks in predicting RBP–RNA interactions and offers a more accurate tool for studying PTBP1 binding specificity.