Lancet2: Improved and accelerated somatic variant calling with joint multi-sample local assembly graphs.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Accurate somatic variant detection is crucial for cancer research and precision medicine.
• Existing short-read sequencing analysis tools face challenges in identifying small variants, particularly insertions and deletions (InDels).

Purpose of the Study:

• To introduce Lancet2, an enhanced open-source somatic variant caller designed for improved small variant detection in short-read sequencing data.
• To optimize somatic variant scoring and visualization for more accurate and efficient analysis.

Main Methods:

• Utilized partial order multiple sequence alignment of haplotype contigs and read re-alignment to best supporting alleles for variant discovery.
• Implemented explainable machine learning models for optimized somatic variant scoring and filtering.
• Integrated Sequence Tube Map for enhanced visualization of variants in graph space.

Main Results:

• Lancet2 demonstrated superior performance in variant calling, particularly for InDels, when benchmarked against leading tools using high-coverage short-read and long-read data.
• Achieved significant runtime improvements (approximately 10x speedup) and reduced memory usage compared to Lancet1.
• Showed at least a 2x speedup with eight or more cores compared to other state-of-the-art callers.

Conclusions:

• Lancet2 provides a highly accurate and efficient solution for somatic variant calling from short-read sequencing data.
• Its advanced methods for variant discovery, scoring, and visualization make it a valuable tool for genomic research and clinical applications.
• The performance enhancements make Lancet2 suitable for large-scale sequencing projects.