Identifying topologically associating domains using differential kernels.

Area of Science:

• Genomics
• Computational Biology
• Molecular Biology

Background:

• Chromatin, a complex of DNA and proteins, regulates gene expression through its three-dimensional (3D) organization.
• High-throughput chromatin conformation capture (Hi-C) techniques generate data to infer 3D chromatin structure.
• Topologically associating domains (TADs) are self-interacting regions within chromatin, crucial for gene regulation, but existing identification algorithms struggle with nested or overlapping TADs.

Purpose of the Study:

• To develop a novel computational method for identifying topologically associating domains (TADs) in Hi-C maps.
• To address limitations of existing algorithms in detecting nested and overlapping TADs.
• To improve the accuracy and consistency of automated TAD identification.

Main Methods:

• Development of KerTAD, a novel method employing a kernel-based technique from computer vision and image processing.
• Benchmarking KerTAD against state-of-the-art TAD identification algorithms using synthetic and experimental Hi-C datasets.
• Evaluation of KerTAD's performance in terms of true positive rates (TPR) and false discovery rates (FDR), and its robustness to noise and sparsity.

Main Results:

• KerTAD accurately identifies nested and overlapping TADs.
• KerTAD demonstrates consistently higher true positive rates (TPR) and lower false discovery rates (FDR) compared to existing methods.
• KerTAD exhibits robustness to increasing noise and sparsity in Hi-C data and consistent TAD identification across experimental replicates.

Conclusions:

• KerTAD offers a significant advancement in automated TAD identification from Hi-C data.
• The method's ability to detect complex TAD structures will enhance the study of gene regulation, enhancer-promoter interactions, and disease mechanisms.
• KerTAD provides a more reliable tool for researchers investigating chromatin organization and its functional consequences.