The National Board of Forensic Medicine annually performs around 5,500 forensic autopsies that are analysed for the presence of medical or other drugs in the body. Data from these cases will contribute to the research now funded by the Swedish Research Council. The researchers will measure the presence of various substances in blood, such as sugars, hormones, signalling molecules and fats, using high-resolution mass spectrometry. An analysis of the total of all the small molecules circulating in the body – the metabolome – may provide a picture of the person’s health status and metabolism at the time of death. Using machine learning and artificial intelligence, researchers will now examine whether they can build models that may eventually help forensic pathologists, police and prosecutors to determine the cause of death more precisely.
Henrik Gréen, professor. Photo credit Emma Busk Winquist
“My hope is that we’ll become much better at determining cause of death and more certain to find murder cases, such as poisonings and similar. We may also be able to find markers for the deadliest conditions, such as sepsis, and markers for chronic heart failure and other conditions,” says Henrik Gréen, professor in the Department of Biomedical and Clinical Sciences at Linköping University and researcher at the National Board of Forensic Medicine.
The contribution from the Swedish Research Council is for a research environment, and the purpose is for researchers to collaborate in larger groups, working towards a common research objective in the long term. In Linköping this will be an interdisciplinary research environment, bringing together experts in forensics, machine learning and AI, systems biology and analytical chemistry. In addition to Henrik Gréen, the group consists of Fredrik Heintz and Oleg Sysoev in the Department of Computer and Information Science, Elin Nyman in the Department of Biomedical Engineering, Johan Dahlén in the Department of Physics, Chemistry and Biology, and Fredrik Kugelberg and Carl Söderberg from the Division for Forensic Genetics and Forensic Toxicology at the National Board of Forensic Medicine.
“We’ve already shown that this works for specific causes of death, so it was more about finding more complex patterns and developing new prediction models, and also about how to implement this new knowledge in our operations. Some of this, such as individual markers, we’ll be able to start using immediately. But if we find more complicated patterns in the metabolome, it’s going to take a bit longer, as this requires building software that can work with such patterns in large-scale routine analysis,” says Henrik Gréen.