Machine learning can help maggots solve crimes
Trained algorithms can identify maggots and read clues like time of death
Maggots don’t have fingers, but they do produce chemical “fingerprints” — a blend of chemicals unique to their species. Different maggot species feed on corpses at particular stages of decay. Forensics teams actually use this information to estimate a person's time of death. The problem is that maggots, which are immature flies, are tricky to identify and rearing them to flyhood is time-consuming and expensive.
University of Albany researchers developed a machine learning technique to rapidly distinguish maggot species by their chemical fingerprints.
Because maggots collected at a crime scene are usually an assortment of many species, the team trained their computer program with mixtures of various combinations of six fly species. The researchers deciphered the chemical signatures of each maggot melange with a technique called mass spectrometry, which identifies chemicals on the basis of size and charge.
After teaching the program to recognize the chemical makeup of each training mixture, the researchers assessed whether it could identify “test” maggot blends. The program accurately identified most of the test samples, suggesting that this technique could help forensic detectives establish time of death, a critical clue for solving crimes. The results are published in a recent issue of Analytical Chemistry.