AI techniques speed up forensic analysis of crucial crime scene larvae

A mass of maggots writhing over a decomposing murder victim isn’t a sight for the squeamish, but for some, it’s evidence. The age and species of a maggot can provide critical information to forensic entomologists investigating murders. (A single wriggling horsefly fly, for example, found on a corpse far from water, gave entomologists in 2022 a key lead in determining where the body came from.) By combing through these fly larvae, investigators can potentially learn when and where a crime took place, whether the body was moved, or whether toxins were involved.

For example, blowflies are among the first insects to colonize corpses; they usually sniff around and lay eggs on a corpse within minutes or even hours. The rate at which maggots (also called larvae) develop depends on heat, humidity, and the species and sex of the insect. To use this evidence, investigators typically must grow larvae to adulthood in the laboratory and then identify them, either visually or by genetic sequence. But what happens if the larvae are dead or missing, there isn’t high-quality DNA, or there isn’t the time (or equipment) to sequence the flies’ genome? “People working in a crime lab simply don’t have the expertise or resources to routinely perform DNA analysis on insect evidence,” says Rabi Musah, a bioorganic chemist at Louisiana State University.

To address these challenges, Musah and other researchers combined machine learning algorithms with methods such as infrared spectroscopy and chemical profiling to quickly identify the species and sex of maggots. Such tools could help experts quickly identify maggots without the larvae’s DNA or without the larvae, just what they leave behind, saving time and money usually spent on sequencing. They could also help investigators take measurements at the crime scene to determine the sex of the larvae.

On supporting science journalism

If you enjoy this article, please consider supporting our award-winning journalism by subscription. 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 metabolome, of insect eggs, larvae and pupae using a type of mass spectrometry, a technique that can separate molecules called metabolites based on their mass and charge. Using this data, she and her team are building a large database of the metabolomes of most of the insects that colonize the decaying remains. His team’s machine learning algorithm, trained on the data, would allow investigators using a mass spectrometer, which is less expensive and much easier to use than a DNA sequencer, to reliably associate a new chemical profile with an insect species in less than five minutes.

A similar approach can work even without the larvae themselves. Sometimes people discover fully decomposed bodies months or years after a murder. By then, the larvae are long gone, Musah says, and the only remaining evidence of the insects is the hard shells of the pupae’s exteriors, tools of metamorphosis left behind once the larvae become adult flies.

It is impossible to identify pupal coats with the naked eye, and in many cases the DNA they contain is too old and degraded to be sequenced. But as Musah’s group reported in a recent Forensic chemistrytheir method – chemical fingerprinting followed by machine-assisted classification – also works with guts. Discovering the chemical profile of the guts can even reveal toxins in the victims’ bodies, as the larvae tend to store them in their pupal coverings. (The rate of molecular degradation could also one day indicate the age of the casings.)

Other groups are also trying to use machine learning to catalog larval visitors to crime scenes: for example, a team of Texas A&M researchers recently developed a method that combines infrared measurements from a handheld device with machine learning to identify the sex of fly larvae.

Male and female larvae develop at different rates and can help investigators determine when they first colonized the remains, but their sexes are impossible to distinguish with the naked eye. To identify the sexes, investigators can crush the larvae and amplify their DNA by PCR, which is time-consuming, renders the larvae useless for further study, and has only an 80 percent chance of working correctly. Texas A&M toxicology graduate student Aidan Holman and his colleagues set out to determine the sex of the larvae without having to crush them.

After 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 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 sex of larvae with more than 90% accuracy. Next, they will collect data on a much wider 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 formally verified, as DNA sequence banks are, so that the results are not later legally overturned. And the broader use of AI in this process can be risky, she adds. “The AI ​​coin flip can become very dangerous in a forensic context, because you can actually cause a miscarriage of justice,” she says. Additionally, she and Musah both emphasize that more research is needed on how other substances in the body might skew molecular markers — and Musah mines data from as large and global a sample of insects as possible to find the markers that remain constant. “Improving and expanding the database involves a never-ending process,” says Musah.

Texas A&M forensic entomologist Jeff Tomberlin, who also was not involved in either project, believes that cutting-edge methods such as machine learning should be integrated into forensic work. But, he notes, their accuracy, precision and potential long-term biases also need to be carefully considered. “We’re in the early stages of applying these methods in this particular area,” he says. “So if you think of it as an arc, we’re at the beginning of the arc.”

It’s time to defend science

If you enjoyed this article, I would like to ask for your support. Scientific American has been defending science and industry for 180 years, and we are currently experiencing perhaps the most critical moment in these two centuries of history.

I was a Scientific American subscriber since the age of 12, and it helped shape the way I see the world. SciAm always educates and delights me, and inspires a sense of respect for our vast and magnificent universe. I hope this is the case for you too.

If you subscribe to Scientific Americanyou help ensure our coverage centers on meaningful research and discoveries; that we have the resources to account for decisions that threaten laboratories across the United States; and that we support budding and working scientists at a time when the value of science itself too often goes unrecognized.

In exchange, you receive essential information, captivating podcasts, brilliant infographics, newsletters not to be missed, unmissable videos, stimulating games and the best writings and reports from the scientific world. You can even offer a subscription to someone.

There has never been a more important time for us to stand up and show why science matters. I hope you will support us in this mission.