Forensic human identification with targeted microbiome markers using nearest neighbor classification.

Area of Science:

• Forensic genetics
• Microbiome research
• Human identity testing

Background:

• The human microbiome presents a promising, yet largely unexplored, resource for forensic genetics.
• Microbiome variability offers potential for advanced human identification beyond traditional methods.
• Predictive accuracy is challenged by microbial distribution variability, necessitating identification of stable microbial signatures.

Purpose of the Study:

• To compare the individualizing power of microbial strain composition versus phylogenetic relatedness for human identification.
• To evaluate two distinct prediction strategies: phylogenetic distance and population genetic diversity measures.
• To determine if microbial signatures are more individualizing than estimates of genetic relatedness between microbial samples.

Main Methods:

• Targeted sequencing of 286 markers across 22 microbial taxa from 51 individuals.
• Sampling across three body sites, with measurements taken in triplicate.
• Utilizing nearest neighbor and reverse nearest neighbor classifiers based on pooled data.

Main Results:

• Microbial diversity-based classification achieved 71% (nearest neighbor) and 78% (reverse nearest neighbor) accuracy.
• Phylogenetic distance-based classification yielded significantly lower accuracy (54% and 63%).
• Diversity-based classification reached 100% accuracy under specific conditions, while phylogenetic methods did not saturate.

Conclusions:

• Microbial strain composition is a more powerful individualizing signature than phylogenetic relatedness.
• Host-environment interactions likely play a role in maintaining stable, individual-specific microbial profiles.
• These findings support the potential of microbiome analysis for advanced forensic identification.