KRLMM: an adaptive genotype calling method for common and low frequency variants

Area of Science:

• Genetics
• Bioinformatics
• Genomic Data Analysis

Background:

• SNP genotyping microarrays are crucial for complex disease research.
• Current algorithms struggle with low-frequency and rare variants due to clustering limitations.
• Accurate genotype calling is essential for variant analysis.

Purpose of the Study:

• To develop KRLMM, a novel method for genotype calling from raw intensity data.
• To address the challenge of accurately identifying low-frequency variants.
• To improve SNP genotyping accuracy in complex disease studies.

Main Methods:

• Developed KRLMM, a genotype calling algorithm.
• Implemented between-sample normalization.
• Allowed variable cluster numbers (k=1, 2, or 3) per SNP, predicted from data.
• Compared KRLMM against GenCall, GenoSNP, Illuminus, and OptiCall using Illumina data and HapMap samples.

Main Results:

• KRLMM demonstrated high overall accuracy (>98%), performing competitively with other leading methods.
• KRLMM, OptiCall, and GenoSNP showed superior accuracy for low-frequency variants compared to GenCall and Illuminus.
• The method successfully converts raw intensities into genotype calls, improving variant detection.

Conclusions:

• Tailored methods like KRLMM (variable clusters) and OptiCall/GenoSNP (using SNP information) enhance accuracy over non-tailored approaches.
• KRLMM offers improved genotype calling for low-frequency variants.
• The KRLMM algorithm is available in the open-source crlmm package via Bioconductor.