AI System 'mHolmes' Uses Microbiome Data to Estimate Time of Death with High Accuracy
Researchers have developed an AI system called mHolmes that estimates the time since death by analyzing microbial changes on cadavers. Published in Nature Communications, the tool achieves an average error of less than two days, outperforming traditional methods that often err by three days or more. The system can also reconstruct microbial history for bodies discovered late or in fragmented states, offering a potential breakthrough for forensic investigations.
Key points
- A new AI model named mHolmes was trained on skin microbiome samples from 34 human cadavers to estimate postmortem intervals.
- The system achieves an average error of less than two days, compared to the ±3 day error typical of previous microbiome-based methods.
- mHolmes can predict microbial changes day by day and reconstruct microbial development from the first week after death, even if data is missing.
- The AI identifies specific bacteria groups associated with decomposition stages, such as Gammaproteobacteria early on and Clostridia during active decay.
- The system can transfer knowledge from one body part to another, which is useful for investigating dismembered or partial remains.
- Researchers note that the small dataset requires further real-world testing and the development of sampling protocols before court use.
Background
This development follows a broader trend in 2026 where AI is being applied to complex biological and forensic challenges. While previous AI advancements focused on economic growth, governance, and safety, this study highlights the potential for AI to enhance precision in scientific fields like forensic science by leveraging biological data.
Why it matters
Accurate estimation of time of death is critical in forensic investigations, especially for cases involving fragmented remains or delayed discovery. mHolmes offers a more precise and flexible tool than traditional methods, potentially improving the reliability of forensic evidence and aiding in the resolution of criminal cases.
What to watch
The next steps involve expanding the dataset, testing mHolmes in various real-world environments, and developing standardized sampling protocols. Researchers aim to validate the system's accuracy across different body parts and conditions before it can be considered for legal use.
Want the full story? Read the original reporting
Read on Gizmodo