An accurate algorithm for the detection of DNA fragments from dilution pool sequencing experiments

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Current high-throughput sequencing technologies produce short reads, limiting the recovery of long-range haplotype information.
• Dilution pool methods allow for the recovery of long DNA fragments from short sequence reads, but require sophisticated computational approaches for fragment identification and haplotype assembly.

Purpose of the Study:

• To develop a computational method for identifying DNA fragments from dilution pool sequencing data.
• To improve the accuracy and contiguity of haplotype assembly using identified DNA fragments.

Main Methods:

• Formulated DNA fragment detection as a genome segmentation problem.
• Developed a dynamic programming algorithm (FragmentCut) to optimize a likelihood function from a generative model.
• Iteratively inferred mean background read depth and number of fragments per pool.

Main Results:

• FragmentCut demonstrated 25-30% greater sensitivity in fragment detection compared to an HMM-based method.
• FragmentCut successfully detected overlapping fragments and assembled haplotypes with improved N50 length and reduced switch and mismatch error rates.
• Validated FragmentCut's accuracy on human fosmid pool and two additional dilution pool datasets.

Conclusions:

• FragmentCut effectively identifies DNA fragments from dilution pool sequencing data.
• The method significantly enhances the accuracy and contiguity of long-read haplotype assembly.
• FragmentCut offers a valuable computational tool for genomic research requiring long-range haplotype information.