A Comparative Study for Identifying the Chromosome-Wide Spatial Clusters from High-Throughput Chromatin Conformation...

Area of Science:

• Genomics
• Computational Biology
• Bioinformatics

Background:

• High-throughput sequencing has advanced genomic annotation, but 3D genome organization remains largely unexplored.
• High-throughput Chromosome Conformation Capture (Hi-C) techniques provide insights into spatial genome arrangements.
• Non-negative Matrix Factorization (NMF) is used to identify spatial clusters in Hi-C data but faces optimization challenges.

Purpose of the Study:

• To develop and evaluate novel optimization algorithms for Non-negative Matrix Factorization (NMF) applied to High-throughput Chromosome Conformation Capture (Hi-C) data.
• To improve the accuracy and efficiency of identifying local spatial clusters of genomic regions from Hi-C datasets.
• To address the challenges of optimizing high-dimensional, non-convex, and constrained objective functions in NMF.

Main Methods:

• Proposed and compared over ten population-based optimization algorithms inspired by in silico evolution.
• Integrated NMF as a local search mechanism within an evolutionary framework.
• Utilized a population of evolving candidates to guide NMF optimization towards global optima.

Main Results:

• The proposed population-based optimization algorithms significantly improved the quality of NMF compared to existing state-of-the-art methods.
• Demonstrated effectiveness and robustness through comprehensive benchmarking on yeast and human chromosome-wide Hi-C contact maps.
• Validated through time complexity, convergence, parameter analyses, biological case studies, and gene ontology similarity analyses.

Conclusions:

• Population-based optimization offers a robust approach to enhance NMF for 3D genome structure analysis using Hi-C data.
• The developed algorithms provide improved identification of spatial genomic clusters, advancing our understanding of genome organization.
• These methods offer a reliable computational tool for analyzing complex Hi-C datasets in genomics research.