Achieving improved accuracy for imputation of ancient DNA

Area of Science:

• Genetics
• Bioinformatics
• Paleogenomics

Background:

• Ancient DNA (aDNA) studies face challenges due to limited sample material.
• Existing genotype imputation tools are often optimized for modern, not ancient, DNA data.
• There is a need for specialized imputation methods to maximize information from aDNA.

Purpose of the Study:

• To investigate the utility of a probabilistic imputation method based on the Li and Stephens model for ancient DNA data.
• To develop and evaluate a novel imputation software, prophaser, tailored for aDNA challenges.
• To compare prophaser's performance against existing imputation pipelines used in the aDNA field.

Main Methods:

• Implementation of prophaser using the Li and Stephens model for haplotype frequency.
• Comparison of prophaser with Beagle-based imputation pipelines.
• Testing on empirical ancient DNA data downsampled to low coverages.
• Evaluation using present-day samples with thinned genotypes.

Main Results:

• Prophaser demonstrates improved accuracy and ability to capture rare variation at lower coverages compared to alternative methods.
• The software is optimized for massive parallel processing and achieves efficient runtimes on GPUs.
• The proposed method effectively addresses the challenges posed by low-quality and limited ancient DNA data.

Conclusions:

• The prophaser imputation method offers significant advantages for analyzing ancient DNA, particularly with limited or low-coverage samples.
• This tool enhances the recovery of genetic information from ancient specimens.
• Prophaser provides a valuable computational resource for paleogenomic research.