Timesweeper: accurately identifying selective sweeps using population genomic time series.

Area of Science:

• Population Genetics
• Evolutionary Biology
• Genomics

Background:

• Identifying selective sweeps, genomic signatures of positive selection, is a persistent challenge in population genetics.
• Existing methods often cannot leverage genomic time-series data due to limitations in historical population sampling.
• Advances in sequencing, particularly ancient DNA and high-throughput methods, now enable serial population sampling.

Purpose of the Study:

• To develop a fast and accurate computational tool for detecting selective sweeps using genomic time-series data.
• To leverage convolutional neural networks for analyzing multiple population samplings over time.
• To improve the resolution and accuracy of identifying targets of positive selection.

Main Methods:

• Timesweeper utilizes a one-dimensional convolutional neural network (CNN) architecture.
• Training data is simulated using demographic models relevant to the specific population.
• The CNN is trained on simulated data to identify polymorphisms under selection in time-series genomic datasets.

Main Results:

• Timesweeper demonstrates high accuracy across various simulated demographic and sampling scenarios.
• The tool precisely identifies selected variants with high resolution.
• It provides more accurate estimations of selection coefficients compared to existing methods.

Conclusions:

• Genomic time-series data significantly enhance the accuracy of natural selection inferences.
• Timesweeper offers a powerful new method for analyzing evolutionary dynamics using serial samples.
• This tool has the potential to advance our understanding of positive selection's role in shaping genomes.