Chromatin loop anchors predict transcript and exon usage.

Area of Science:

• Genomics and Bioinformatics
• Molecular Biology
• Computational Biology

Background:

• High-throughput sequencing data like RNA-seq and ChIP-seq are established for gene expression prediction.
• Chromatin Interaction Analysis with Paired End Tag sequencing (ChIA-PET) identifies genome-wide chromatin loops but hasn't been used for predictive tasks.
• Understanding gene regulation requires integrating diverse genomic data types.

Purpose of the Study:

• To investigate the predictive power of ChIA-PET data for transcript and exon usage.
• To develop and validate machine learning models integrating ChIA-PET and other genomic data.
• To identify key genomic features driving transcript and exon usage predictions.

Main Methods:

• Development of Gradient Boosting Trees models using integrated datasets from three cell lines (GM12878, HeLaS3, K562).
• Inclusion of transcription factor data and ChIA-PET data in model training.
• Rigorous model validation using 10-fold cross-validation, chromosome-split validation, and cross-cell validation.

Main Results:

• Accurate prediction of transcript usage (accuracy ≥ 0.7512) and splicing-derived exon usage (accuracy ≥ 0.7459) across all cell lines and validation strategies.
• Identification of RNA Polymerase II ChIA-PET as a highly important predictive feature.
• Demonstration that chromatin loop anchors are significant predictors of both transcript and exon usage.

Conclusions:

• ChIA-PET data significantly enhances the prediction of gene transcript and exon usage.
• Machine learning models can effectively leverage ChIA-PET information for gene expression and splicing predictions.
• Chromatin loop structures play a crucial role in regulating gene expression and alternative splicing patterns.