Gaussian process test for high-throughput sequencing time series: application to experimental evolution

Area of Science:

• Genomics
• Population Genetics
• Bioinformatics

Background:

• High-throughput sequencing (HTS) enables detailed genomic monitoring over time.
• Identifying significant changes in genomic feature abundance is crucial for understanding biological processes like selection pressures.
• Existing methods struggle to incorporate intermediate time points, replicate experiments, and HTS-specific uncertainties like sequencing depth.

Purpose of the Study:

• To develop a novel statistical model for analyzing time-series HTS data.
• To accurately rank genomic features exhibiting significant non-random variation in abundance over time.
• To improve the detection of selection pressures in population genetics studies.

Main Methods:

• The beta-binomial Gaussian process model is proposed.
• The beta-binomial model accounts for uncertainty from finite sequencing depth.
• A Gaussian process model is integrated to handle time-series data.

Main Results:

• The beta-binomial Gaussian process model demonstrates superior average precision in identifying selected alleles compared to classical testing in simulations.
• Simulations explored the impact of experimental design choices on model performance.
• The model was successfully applied to real data from a Drosophila experimental evolution experiment.

Conclusions:

• The beta-binomial Gaussian process model offers a robust approach for analyzing time-series HTS data.
• This method enhances the ability to detect significant genomic changes and selection pressures.
• The developed R software package facilitates the application of this advanced statistical method.