They then analysed the bacterial samples using next-generation metagenomic DNA sequencing to determine the identity and abundance of all microbes present.
With this very large data set, the researchers used a machine learning approach to scrutinise the bacterial communities and how they change over time as the bodies decomposed. Through iterative testing and tweaking of their computational tools, the researchers built a statistical model that predicts the postmortem interval of unknown samples to within 55 accumulated degree-days, or about two days in summertime.
This degree of accuracy holds through several weeks of decomposition, a substantial improvement over presently available methods. “Our approach had the benefit of sampling the same cadavers repeatedly as they decomposed and we think that this really added to the ability of our machine learning approach to see through all of the massive amount of noisy data and detect the underlying patterns,” said Lents.
“While we consider this a pilot study, it is a very promising proof-of-concept, and I think that microbiome-based approaches will eventually become the standard method of determining the time since death for bodies that are discovered after some time of decomposition,” Lents said.