Accounting for bias from sequencing error in population genetic estimates

Area of Science:

• Genomics
• Population Genetics
• Bioinformatics

Background:

• Sequencing errors pose a significant challenge for population genetic analyses, particularly with low-coverage and single-pass sequencing data.
• Parameter estimates in population genetics can be severely biased when the level of polymorphism is low compared to sequencing error rates.
• Current methods using arbitrary quality score cutoffs are problematic, especially with newer sequencing technologies exhibiting diverse quality score distributions.

Purpose of the Study:

• To address the challenge of sequencing errors in population genetic studies.
• To evaluate the impact of quality score cutoffs on parameter estimates.
• To propose a method for identifying biased thresholds and suggest bias-reducing alternatives.

Main Methods:

• Analysis of sequencing error rates in low-coverage and single-pass reads.
• Evaluation of bias in population genetic parameter estimates under varying signal-to-noise ratios.
• Development of a rule of thumb to assess the impact of quality score cutoffs.
• Exploration of alternative bioinformatic approaches to mitigate bias.

Main Results:

• Arbitrary quality score cutoffs can lead to significant bias in population genetic parameter estimates.
• The severity of bias is exacerbated when polymorphism levels are low relative to sequencing error.
• Newer sequencing technologies present unique challenges due to their distinct quality score distributions.
• A practical rule of thumb is proposed to help researchers identify potentially biased thresholds.

Conclusions:

• Standard quality score cutoff methods are insufficient for accurate population genetic analyses with noisy sequence data.
• Researchers must be cautious when selecting quality thresholds to avoid biased estimates.
• Alternative analytical strategies are necessary to improve the reliability of population genetic inferences from low-coverage sequencing data.