Abstract:
|
Metagenomic analysis is a promising alternative to traditional forensic methods when analyzing trace evidence found at the crime scene. Microbial samples found on evidence usually contain microbes from multiple contributors. However, traditional clustering methods such as k-means, multivariate analysis, and hierarchical clustering may have difficulty correctly identifying the true relationships between evidence and contributors when correlations between contributors are high and proportions are small. We propose a hierarchical method with adjustment for dendogram structure before inclusion of any evidence for use in predicting significant contributors. Analysis of simulated evidence samples show promising results.
|