A graph-based algorithm for estimating clonal haplotypes of tumor sample from sequencing data.

Area of Science:

• Bioinformatics
• Computational Genomics
• Cancer Research

Background:

• Haplotype phasing is crucial in bioinformatics, especially for cancer genomics.
• Reconstructing tumor clonal haplotypes aids understanding of clonal architecture for diagnosis and treatment.
• Existing algorithms struggle with sequencing data complexity, leading to high computational costs and inaccuracies.

Purpose of the Study:

• To develop an efficient computational pipeline for phasing cancer clonal haplotypes.
• To address the limitations of existing algorithms in terms of time and space complexity.
• To improve the accuracy and practicality of reconstructing clonal haplotypes from sequencing data.

Main Methods:

• Proposed MixSubHap, a graph-based computational pipeline for cancer sequencing data.
• Implemented three bounding strategies to reduce solution space and filter false positives.
• Utilized variant allelic frequencies for global clonal structure estimation, greedy extension for contig linkage, and read-depth stripping for filtering.

Main Results:

• MixSubHap identifies approximately 90% of preset clonal haplotypes across various simulations.
• The pipeline demonstrates robustness with decreasing mutation rates, achieving contig lengths up to 10kbps.
• Mutation-to-haplotype assignment accuracy remains above 60% on average.

Conclusions:

• MixSubHap is a practical and efficient algorithm for reconstructing cancer clonal haplotypes.
• The method offers significant improvements in accuracy and computational performance.
• The developed pipeline provides valuable insights for cancer genomics and potential clinical applications.