GBScleanR: robust genotyping error correction using a hidden Markov model with error pattern recognition.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Reduced-representation sequencing (RRS) offers efficient genotyping but yields noisy data with significant errors.
• Existing hidden Markov model (HMM) based error correction methods assume uniform error rates, failing to account for biases.
• Allele read ratio bias arises from uneven fragment amplification and read mismapping in RRS data.

Purpose of the Study:

• To introduce GBScleanR, a novel error correction tool for RRS genotype data.
• To enhance genotype accuracy by incorporating marker-specific error rates into HMMs.
• To provide a robust and precise method for correcting errors in noisy RRS data.

Main Methods:

• Development of GBScleanR, an error correction tool utilizing HMMs.
• Incorporation of marker-specific error rates to address biases in RRS data.
• Validation using both simulation and real-world RRS genotype datasets.

Main Results:

• GBScleanR significantly improves genotype accuracy, achieving up to a 25 percentage point increase compared to existing tools on simulated data.
• The tool demonstrates reliable genotype estimation on real-world datasets, even with error-prone markers.
• Marker-specific error rate incorporation leads to more precise error correction.

Conclusions:

• GBScleanR offers a robust and precise solution for correcting errors in RRS-based genotype data.
• The method effectively handles marker-specific biases, outperforming existing tools.
• This tool enhances the reliability of genotype data derived from cost-effective sequencing platforms.