A mass of writhing maggots on a decomposing murder victim is not a sight for the squeamish, but for some, it is evidence. A maggot’s age and species can give essential information to forensic entomologists investigating murders. (A single wriggling horse fly maggot, for instance, found on a dead body far from water, gave entomologists in 2022 a key lead to where the body came from.) Combing through these fly larvae, investigators can potentially learn when and where a crime happened, whether the body has been moved or whether toxins were involved.
For example, blowflies are among the earliest insect colonizers of corpses; they typically sniff out and lay eggs on a dead body within minutes to hours. How fast the maggots (also called larvae) develop depends on heat, humidity, and the insect’s species and sex. To use this evidence, investigators typically must grow the larvae until adulthood in a laboratory setting and then identify them, either visually or by genetic sequence. But what if the larvae are dead or missing, there’s no high-quality DNA or there isn’t the time—or equipment—to sequence the flies’ genomes? “People in a crime lab simply do not have the resident expertise or the resources to be able to routinely conduct DNA analysis on insect evidence,” says Rabi Musah, a bioorganic chemist at Louisiana State University.
To tackle these challenges, Musah and other researchers have combined machine-learning algorithms with methods such as infrared spectroscopy and chemical profiling to quickly pinpoint maggots’ species and sex. Such tools could help experts rapidly identify the maggots without the larvae’s DNA or without the larvae altogether, only what they leave behind—saving time and money usually spent in sequencing. They could also help investigators take measurements at the scene of the crime itself to determine larval sex.
On supporting science journalism
If you're enjoying this article, consider supporting our award-winning journalism by subscribing. By purchasing a subscription you are helping to ensure the future of impactful stories about the discoveries and ideas shaping our world today.
Musah measured the chemical profiles, called the metabolome, of insect eggs, larvae and pupae using a type of mass spectrometry—a technique that can tease apart molecules called metabolites based on their mass and charge. With these data, she and her team are building a large metabolome database for most of the insects that colonize decomposing remains. Her team’s machine-learning algorithm trained on the data would let investigators using a mass spectrometer, which is less expensive and much easier to use than a DNA sequencer, reliably match a new chemical profile to an insect species in less than five minutes.
A similar approach can work even without the larvae themselves. Sometimes people come across fully decomposed bodies many months or years after a murder. By that time the larvae are long gone, Musah says, and the only remaining insect evidence is the hard shell-like exteriors of the pupae, tools of metamorphosis discarded after the larvae become adult flies.
It’s impossible to identify pupae coverings with the naked eye, and in many cases, the DNA contained within them is too old and degraded for sequencing. But as Musah’s group reported in a recent paper in Forensic Chemistry, their method—of chemical fingerprinting followed by machine-assisted classification—works with casings, too. Finding the chemical profile of casings can even reveal toxins in the victims’ bodies because the larvae tend to store them in their pupal coverings. (The rate of molecular breakdown might also someday point to the casings’ ages.)
Other groups are also trying to use machine learning to catalog crime scenes’ larval visitors: for instance, a team of Texas A&M researchers recently developed a method that combines infrared measurements from a handheld device with machine learning to identify blowfly larvae’s sexes.
Male and female larvae develop at different speeds and can help investigators pinpoint when they first colonized remains, but their sexes are indistinguishable by eye. To identify sexes, investigators can crush the larvae and amplify their DNA using PCR, which is time-consuming, renders the larvae useless for any further studies, and has only an 80 percent chance of working correctly. Aidan Holman, a Texas A&M toxicology graduate student, and his colleagues set out to find larvae’s sexes without having to mash them up.
After first rearing the male and female larvae separately, Holman’s group used a handheld infrared spectroscopy device to “zap” them and measure the light released. The proteins, fats and other molecules that make up the larvae scatter the light in unique ways, generating a specific “spectral signature” based on sex. The researchers then trained a machine-learning model on this spectral data and found that it could predict the larvae’s sex with more than 90 percent accuracy. Next, they will collect data from a much bigger selection of flies to train their model.
Murdoch University forensic entomologist Paola Magni, who is not involved in either project, emphasizes that these machine-learning databases will need to be officially vetted, as DNA sequence banks are, so results aren’t later overturned legally. And use of AI more broadly in this process can be risky, she adds. “The flip of the coin of artificial intelligence can become very dangerous in a forensic context because you can really cause a miscarriage of justice,” she says. Plus, she and Musah both highlight that more research is needed into how other substances in the body might skew molecular markers—and Musah is pulling data from across as large and global an insect sample as possible to find the markers that remain constant. “The enhancement and expansion of the database involves a never-ending process,” Musah says.
Texas A&M forensic entomologist Jeff Tomberlin, who was also not involved in either project, believes that cutting-edge methods like machine learning should be integrated into forensic case work. But, he notes, their long-term accuracy, precision and potential biases need to be carefully studied as well. “We’re in the infancy of applying these methods in this particular realm,” he says. “So if you think of it like an arc, we’re at the beginning of the arc.”