BMI-CNV: a Bayesian framework for multiple genotyping platforms detection of copy number variants

Area of Science:

• Genomics
• Bioinformatics
• Cancer Research

Background:

• Whole-exome sequencing (WES) offers high-resolution copy number variant (CNV) detection but is limited to protein-coding regions.
• Intergenic and intragenic variants are often missed, and single-sample methods have high false discovery rates.
• Integrating data from multiple genotyping platforms and samples is crucial for robust CNV detection.

Purpose of the Study:

• To develop an integrative method for improved CNV detection by combining whole-exome sequencing and microarray data.
• To address the limitations of WES and single-sample analyses in CNV identification.
• To identify shared CNV regions across multiple samples and platforms.

Main Methods:

• Developed BMI-CNV (Bayesian Multisample and Integrative CNV) profiling method.
• Integrated whole-exome sequencing and microarray data.
• Employed a Bayesian probit stick-breaking process model and Gaussian Mixture model for multisample integration and shared CNV region identification.

Main Results:

• BMI-CNV demonstrated superior accuracy compared to existing methods in simulations.
• Accurately detected common variants in 1000 Genomes and HapMap data.
• Significantly expanded the detection spectrum beyond whole-exome sequencing's capabilities.
• Identified potential lung cancer risk CNV candidates in TRICL data (17q11.2, 1p36.12, 8q23.1, 5q22.2).

Conclusions:

• BMI-CNV effectively integrates diverse genomic data for comprehensive CNV profiling.
• The method enhances CNV detection accuracy and broadens genomic coverage.
• BMI-CNV is a valuable tool for genetic studies, including cancer risk variant identification.