A new diagnostic method, developed by researchers at Unicamp, makes it possible to quickly and simply investigate a range of diseases and manifestations that present themselves at a metabolic level. The invention called “molecular selector” combines biochemical analyzes and artificial intelligence – two renowned and different analytical techniques – to create an exclusive platform for examinations such as the detection of changes caused by viruses, bacteria or fungi.
Once trained, the algorithm is capable of automatically identifying, within minutes, biomarkers associated with serious, neglected diseases or those that do not have specific protocols, helping doctors and specialists in decision-making. According to the scientists, the invention can also be applied to other laboratory, environmental and bromatological analyses.
Technology couples metabolomics data with a machine learning classification algorithm (machine learning, in English) to investigate, cross-reference and identify patterns and sets of characteristics that portray a certain condition of interest in an organism. The research involved scientists from the Complex Data Inference Laboratory (Recod), from the Computing Institute (IC), and the Innovare Biomarkers Laboratory, from the Faculty of Pharmaceutical Sciences (FCF), from Unicamp.
The invention was protected through a patent and a computer program with support from Inova Unicamp and has the potential to benefit the pharmaceutical and cosmetics industries, among others. In the agricultural sector, for example, it is possible to work on detecting elements that cause putrefaction or that compromise food, such as the presence of Salmonella, ensuring the quality of the products consumed.
“Any problem that manifests itself at a metabolic level and causes molecular interference or differentiation is likely to have biomarkers classified by our molecular selection platform, from diseases such as Covid-19 to food adulteration or rotting”, summarizes Rodrigo Catharino, director of the Faculty of Pharmaceutical Sciences (FCF), from Unicamp.
Another advantage described by the team is the ability to refine results, increasing the sensitivity of the algorithm and the success rate. “The platform can adapt to conditions that change over time, as is the case with viral infections such as Covid-19. By providing new samples from patients, with and without manifestations of the disease, the method can be improved for new scenarios”, explains Anderson Rocha, Director of the Computing Institute (IC).
According to the professor, it is also possible to identify risks of complications based on biomarkers of contamination and other infections that can worsen the patient's general condition. Additionally, once the machine learns what to look for, it can diagnose more than one disease or condition from the same sample taken, reducing laboratory costs.
Check out full article published on the Unicamp Innovation Agency website.
