Detecting genomic deletions from high-throughput sequence data with unsupervised learning.

Area of Science:

• Genomics
• Bioinformatics
• Molecular Biology

Background:

• Structural variations (SVs), including deletions, are significant genetic variations impacting DNA sequences.
• High-throughput sequencing data analysis commonly utilizes discordant read-pairs, read depth, and split reads for SV detection.
• Existing SV detection tools often rely on one or multiple signal types, with varying performance.

Purpose of the Study:

• To develop a novel computational method, EigenDel, for detecting germline submicroscopic genomic deletions.
• To evaluate the performance of EigenDel, particularly on low-coverage sequencing data.

Main Methods:

• EigenDel integrates discordant read-pairs and clipped reads to identify initial deletion candidates.
• Unsupervised learning methods are employed for clustering similar deletion candidates.
• A specialized calling approach is used to identify true deletions within each cluster.

Main Results:

• EigenDel demonstrates superior performance compared to other major SV detection methods.
• The method effectively balances accuracy and sensitivity in deletion detection.
• EigenDel shows a reduction in bias for germline deletion identification.

Conclusions:

• EigenDel offers an improved capability for detecting germline genomic deletions.
• The method provides a more accurate and sensitive approach to analyzing structural variations.
• EigenDel is available as open-source software for broader research application.