PhredEM: a phred-score-informed genotype-calling approach for next-generation sequencing studies

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Accurate genotype calling is crucial for next-generation sequencing (NGS) data analysis.
• Estimating base-calling error rates is critical for reliable genotype calls.
• Current methods using Phred scores or ignoring them have limitations, including data loss or error introduction.

Purpose of the Study:

• To develop a novel likelihood-based genotype-calling approach, PhredEM.
• To accurately estimate per-base error rates by incorporating Phred scores via logistic regression.
• To improve genotype calling robustness and applicability in NGS studies.

Main Methods:

• PhredEM utilizes an expectation-maximization (EM) algorithm for consistent parameter estimation.
• A logistic regression model incorporates Phred scores to estimate error rates.
• A screening algorithm identifies monomorphic loci to optimize computational efficiency.

Main Results:

• PhredEM demonstrated superior performance compared to GATK and SeqEM on simulated and real sequencing data.
• The approach effectively estimates genotype frequencies and logistic regression parameters.
• PhredEM integrates with linkage-disequilibrium methods like Beagle for enhanced genotype calling.

Conclusions:

• PhredEM offers an improved, robust, and widely applicable genotype-calling method for NGS.
• The software is freely available, promoting its adoption in genomic research.
• This approach enhances the accuracy of downstream analyses in various NGS studies.