Traditional alignment of next-generation sequencing (NGS) data is affected by reference bias, whereby alleles that differ from the reference genome align less efficiently due to the limited representation of variation in a linear haploid reference. Graph-based pangenome references offer an alternative approach, representing genomic regions as networks of sequence variation that enable reads to align along multiple paths within the same locus. Although whole-genome pangenomes remain under development due to the extensive effort required to catalog global variation, forensic marker sets are well characterized, making them suitable targets for graph-based approaches.

This study evaluated whether graph-based alignment could improve heterozygote balance and reduce misalignment noise in microhaplotypes. A total of 35 reference samples, previously amplified using panels of 43 and 60 microhaplotypes, were sequenced on the Ion S5 platform. Sequence data were aligned using the standard linear aligner TMAP and the graph-based aligner vg giraffe. Given that vg giraffe is optimized for Illumina data, where indel error rates are low, an additional analysis incorporated indels into the graph outside allele-defining SNP regions to better accommodate Ion Torrent error profiles.

Alignment performance varied substantially by locus. At some loci, use of vg giraffe increased average heterozygote balance by up to 24% and reduced noise by up to 5%. Conversely, other loci showed decreases in heterozygote balance of up to 9% and increases in noise of up to 3%. Considering loci with ≥1% change, 15 loci exhibited reduced heterozygote imbalance and 13 showed reduced noise, while 11 loci demonstrated decreased balance and 4 increased noise.

Incorporation of indel variation into the graph improved alignment performance at select loci but negatively affected others, indicating that graph design is critical to performance. Overall, graph-based alignment using vg giraffe shows promise for improving microhaplotype analysis, though further optimization of locus-specific graph structures is necessary to achieve consistent benefits across marker sets.

Authors

• Zack Beyerl (Bode Technology, United States of America)
• Robert Lagacé (Thermo Fisher Scientific, United States of America)
• Daniele Podini (The George Washington University, United States of America)