Researchers are working on developing machine-learning algorithms to spot bits of DNA that DNA synthesis companies have been asked to create that could be dangerous, Nature News reports.
It notes that researchers routinely turn to such services, but that there have been concerns that individuals with more nefarious plans could, too. According to Nature News, the US Intelligence Advanced Research Projects Agency started a program in 2016 to design algorithms to spot sequences that encode viruses or toxins or other threats — previous efforts, it notes, have been limited by a need for exact matches and by database errors — so they can then ensure the customer ordering them is a legitimate researcher, and not a bioterrorist.
Nature News reports that researchers like those at Battelle or the University of Virginia are making headway in this effort and have presented their work at American Society for Microbiology Biothreats meeting. For instance, it says UVa's Andrew Warren and his team are developing an algorithm that, by analyzing microbial DNA, learns what sort of traits are common to DNA sequences of toxins and pathogens, while a Battelle team is similarly analyzing pathogens so they can be quickly recognized.