RVD2: an ultra-sensitive variant detection model for low-depth heterogeneous next-generation sequencing data

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Next-generation sequencing (NGS) is vital for clinical diagnostics.
• Clinical samples often exhibit genomic heterogeneity, complicating variant detection.
• A need exists for robust variant calling algorithms for low-frequency polymorphisms in mixed samples.

Purpose of the Study:

• To develop and validate a novel variant calling algorithm for heterogeneous samples.
• To improve accuracy, sensitivity, and specificity in low-frequency variant detection.
• To address limitations of current variant callers in complex clinical specimens.

Main Methods:

• Development of a hierarchical Bayesian model for allele frequency estimation.
• Application of the algorithm to simulate and real-world heterogeneous genomic data.
• Comparative analysis against existing variant calling classifiers.

Main Results:

• The novel algorithm demonstrates improved sensitivity and specificity across various read depths and minor allele fractions.
• Successfully identified 15 mutated loci in the PAXP1 gene from a breast ductal carcinoma tumor sample.
• Detected two potential loss-of-heterozygosity events within the identified mutated loci.

Conclusions:

• The developed hierarchical Bayesian model offers a significant advancement in variant calling for heterogeneous clinical samples.
• The algorithm provides a more reliable tool for low-frequency polymorphism detection in NGS-based diagnostics.
• This method has direct implications for improving the accuracy of cancer genomics and personalized medicine.