A random forest-based framework for genotyping and accuracy assessment of copy number variations

Area of Science:

• Genomics
• Bioinformatics
• Human Genetics

Background:

• Accurate detection of copy number variations (CNVs) is crucial for understanding genetic contributions to human diseases.
• Current methods for CNV identification from whole genome sequencing (WGS) data face challenges in precision and error rates.
• Identifying small CNVs (≤1 kb) and ensuring accurate genotyping remain significant hurdles in genomic research.

Purpose of the Study:

• To introduce CNV-JACG, a novel framework for judging the accuracy of CNVs and genotyping.
• To enhance the precise identification of CNVs from paired-end WGS data.
• To provide a reliable tool for assessing CNV accuracy and aiding in the discovery of missing heritability.

Main Methods:

• Development of CNV-JACG, a framework utilizing a random forest model.
• Training the model on 21 distinctive features related to CNV regions and breakpoints.
• Validation using diverse datasets including the 1000 Genomes Project, Genome in a Bottle Consortium, and technical replicates.

Main Results:

• CNV-JACG demonstrates superior sensitivity compared to the SV² genotyping method, especially for small CNVs (≤1 kb).
• The framework exhibits improved performance in reducing Mendelian inconsistencies within trio families.
• CNV-JACG shows higher concordance rates between technical replicates, indicating enhanced reliability.

Conclusions:

• CNV-JACG offers a robust solution for accurate CNV detection and genotyping from paired-end WGS data.
• The framework's superior performance, particularly for small CNVs, addresses a critical gap in current genomic analysis tools.
• CNV-JACG is poised to become a valuable asset in uncovering genetic factors, including missing heritability, associated with human diseases.