Duphold: scalable, depth-based annotation and curation of high-confidence structural variant calls

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Current structural variant (SV) detection methods primarily rely on read-pair and split-read alignments.
• These methods often do not integrate sequence depth information, a crucial factor for copy number variant (CNV) analysis.
• Lack of depth integration can limit the confidence and accuracy of SV calls, especially in large-scale projects.

Purpose of the Study:

• To introduce duphold, a novel computational method for annotating SV calls with sequence depth information.
• To enhance the accuracy and specificity of SV detection by leveraging sequence coverage patterns.
• To provide an efficient tool for evaluating SVs in large, multi-sample sequencing datasets.

Main Methods:

• Duphold analyzes sequence depth changes across putative structural variant breakpoints.
• It identifies both the magnitude of depth change and the rapidity of this change relative to flanking regions.
• The method employs an algorithm designed for near-constant runtime, irrespective of the number of variants analyzed.

Main Results:

• Filtering SV calls based on duphold annotations significantly improves call specificity.
• Duphold effectively integrates sequence depth data to support or refute predicted copy number alterations.
• The tool demonstrates efficiency, making it suitable for large-scale genomic projects involving numerous samples.

Conclusions:

• Duphold provides a valuable method for enhancing the reliability of structural variant detection.
• The integration of sequence depth information via duphold improves the accuracy of copy number variant identification.
• Duphold is compatible with both short-read and long-read sequencing data, offering broad applicability.