Paper List

期刊: ArXiv Preprint
发布日期: 2026-03-12
BioinformaticsGenomics

A Multi-Label Temporal Convolutional Framework for Transcription Factor Binding Characterization

Unknown

P. Demurtas, F. Zanchetta, G. Perini, R. Fioresi
Figure
Figure
Figure
Figure
Figure

30秒速读

IN SHORT: This paper addresses the critical limitation of existing TF binding prediction methods that treat transcription factors as independent entities, failing to capture their cooperative regulatory mechanisms through multi-label classification.

核心创新

  • Methodology First application of Temporal Convolutional Networks (TCNs) to multi-label transcription factor binding prediction, enabling simultaneous prediction of multiple TF binding events from DNA sequences.
  • Methodology Development of three multi-label datasets (D-5TF-3CL, D-7TF-4CL, H-M-E2F) from ENCODE ChIP-seq data, specifically designed to study TF cooperativity.
  • Biology Demonstration that deep learning models can learn biologically meaningful TF correlations and cooperative patterns directly from DNA sequence data, revealing both known and novel TF interactions.

主要结论

  • TCN-based models significantly outperform RNN baselines in multi-label TF prediction, achieving average F1-score improvements of +0.17 to +0.26 across datasets (p<0.05).
  • The model captures biologically relevant TF correlations, with TCN achieving AP scores of 0.73±0.01 on the H-M-E2F dataset compared to 0.52±0.00 for RNN baselines.
  • TCNs demonstrate robust performance even with limited data, maintaining AP >0.7 on 152 out of 165 binary classification datasets despite moderate correlation (Pearson r=0.61) between performance and dataset size.
研究空白： Current deep learning approaches for DNA-binding recognition focus exclusively on binary classification tasks predicting single TF binding, ignoring the combinatorial nature of transcriptional regulation where multiple TFs cooperate through complex interactions.

摘要: Transcription factors (TFs) regulate gene expression through complex and cooperative mechanisms. While many TFs act together, the logic underlying TFs binding and their interactions is not fully understood yet. Most current approaches for TF binding site prediction focus on individual TFs and binary classification tasks, without a full analysis of the possible interactions among various TFs. In this paper we investigate DNA TF binding site recognition as a multi-label classification problem, achieving reliable predictions for multiple TFs on DNA sequences retrieved in public repositories. Our deep learning models are based on Temporal Convolutional Networks (TCNs), which are able to predict multiple TF binding profiles, capturing correlations among TFs and their cooperative regulatory mechanisms. Our results suggest that multi-label learning leading to reliable predictive performances can reveal biologically meaningful motifs and co-binding patterns consistent with known TF interactions, while also suggesting novel relationships and cooperation among TFs.