Supervised machine learning enables forensic microbiome source attribution in soil and touch samples with evidence‐type‐specific relic DNA effects

Microbial trace evidence is increasingly proposed for forensic source attribution, yet most studies lack performance benchmarks across evidence types and rarely test robustness across sample preprocessing conditions. Here, we evaluated supervised machine learning models for attributing (i) mock soil trace evidence to four source locations and (ii) human touch trace evidence to the correct donor using paired samples collected under controlled transfer conditions. 16S rRNA amplicon profiles were used to assess microbial community diversity, and extreme gradient boosting (XGBoost) classifiers were tested under within-condition and cross-condition train/test schemes. We further tested how extracellular DNA (i.e., relic DNA) depletion altered microbiome diversity and attribution performance. For soil trace evidence, microbial community structure was strongly driven by location, while relic DNA depletion and sample type had smaller, location-specific effects. Classifiers achieved high location attribution performance, with sink accuracies exceeding 84% and reaching 97.9%, indicating that environmental microbial signatures were robust across preprocessing enzymatic treatment conditions. For human touch evidence, donor attribution was feasible but weaker and more condition sensitive. The best performance was observed for control-trained models predicting control PVC sinks, where top-3 accuracy reached 86%, but declined to 68% for relic-DNA-depleted sinks under the same training scheme, indicating that relic DNA depletion can reduce attribution performance for lower-complexity touch evidence. These findings establish a benchmark for supervised microbial trace evidence source attribution across soil and human touch samples while showing that performance depends on evidence type and preprocessing condition and does not support routine relic DNA depletion, particularly for touch evidence.

Authors

Institutions

Publication Details

Journal
Journal of Forensic Sciences
Published
2026-11-16
DOI
https://doi.org/10.1111/1556-4029.70504
Primary Topic
Forensic and Genetic Research
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Supervised machine learning enables forensic microbiome source attribution in soil and touch samples with evidence‐type‐specific relic DNA effects

Zachary M. Burcham, Alison Buchan, Giovanna M. Vidoli, Emily Cantrell et al.
Journal of Forensic Sciences
Forensic and Genetic Research
article

Supervised machine learning enables forensic microbiome source attribution in soil and touch samples with evidence‐type‐specific relic DNA effects

Zachary M. Burcham, Alison Buchan, Giovanna M. Vidoli, Emily Cantrell, Reese J Saho, Savannah Truan‐Walker, Julia Erie Tait
article en

Abstract

Microbial trace evidence is increasingly proposed for forensic source attribution, yet most studies lack performance benchmarks across evidence types and rarely test robustness across sample preprocessing conditions. Here, we evaluated supervised machine learning models for attributing (i) mock soil trace evidence to four source locations and (ii) human touch trace evidence to the correct donor using paired samples collected under controlled transfer conditions. 16S rRNA amplicon profiles were used to assess microbial community diversity, and extreme gradient boosting (XGBoost) classifiers were tested under within-condition and cross-condition train/test schemes. We further tested how extracellular DNA (i.e., relic DNA) depletion altered microbiome diversity and attribution performance. For soil trace evidence, microbial community structure was strongly driven by location, while relic DNA depletion and sample type had smaller, location-specific effects. Classifiers achieved high location attribution performance, with sink accuracies exceeding 84% and reaching 97.9%, indicating that environmental microbial signatures were robust across preprocessing enzymatic treatment conditions. For human touch evidence, donor attribution was feasible but weaker and more condition sensitive. The best performance was observed for control-trained models predicting control PVC sinks, where top-3 accuracy reached 86%, but declined to 68% for relic-DNA-depleted sinks under the same training scheme, indicating that relic DNA depletion can reduce attribution performance for lower-complexity touch evidence. These findings establish a benchmark for supervised microbial trace evidence source attribution across soil and human touch samples while showing that performance depends on evidence type and preprocessing condition and does not support routine relic DNA depletion, particularly for touch evidence.

Journal of Forensic Sciences
University of Tennessee at Knoxville (US)
Forensic and Genetic Research
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon