Preparing for shotgun sequencing in forensic genetics – Benchmarking of tools for read mapping, genotype calling, and imputation

Shotgun sequencing has emerged as a powerful tool in forensic genetics, as it allows for comprehensive genetic profiles to be generated from highly degraded DNA. The method enables simultaneous access to a wide range of markers, thereby supporting applications such as human identification (HID), DNA intelligence, and forensic investigative genetic genealogy (FIGG). However, the accuracy and utility of shotgun sequencing data are highly dependent on the bioinformatic analysis. In this study, we benchmarked widely used bioinformatic tools using shotgun sequencing data from both high-quality (blood and buccal) and low-quality (hair) forensic samples. Specifically, we evaluated five alignment algorithms (Bowtie2, BWA-ALN, BWA-MEM, CLC, and CLC LightSpeed), four genotype calling methods (ANGSD, ATLAS, GATK HaplotypeCaller, and a custom rule-based approach), and three imputation methods (Beagle4.1, Beagle5.4, and GLIMPSE2). All investigated tools were found to be suitable for analysing high-quality reference samples. However, their performance varied significantly when applied to low-quality (hair) forensic samples. The combination of BWA-MEM, ANGSD, or GATK HaplotypeCaller, and imputation with GLIMPSE2 produced the lowest degree of discordance. The work presented here emphasises the importance of informed bioinformatic tool selection and optimisation, and it provides practical recommendations for analysing shotgun sequencing data in forensic genetics.