apart. They found eight bacterial genetic sequences that were more likely to be present in the guts of children with autism than in their non-autistic siblings, and three sequences that were less likely. A follow-up study releasing later in 2025 found further interesting results linked to metabolites, small molecules produced during metabolism. These new findings are exciting because understanding the specific metabolic pathways altered in developmental and neurological disorders could pave the way for novel therapies targeting the gut microbiome. To uncover patterns in these vast datasets, David turns to artificial intelligence. Her lab trains machine learning models on large collections of microbial sequences to detect subtle associations between microbial communities and human health. This data-driven approach helps researchers classify conditions such as inflammatory bowel disease or colorectal cancer with greater accuracy. Now, David is extending her AI work from the human body to the deep sea. With a $540K NSF grant, her team is applying deep learning to analyze methane seeps, areas where gas escapes from the ocean floor, off the coasts of Oregon and Washington. These habitats are home to methane-consuming microbes that play a critical role in regulating the global carbon and nitrogen cycles, yet many of the genes that drive these processes remain unidentified. To address this, David’s team is developing two deep learning models: one to group genes based on how they appear together in microbial communities, and another to use generative AI to predict gene functions from protein sequences and scientific literature. Together, these tools could dramatically improve our understanding of how microbial life shapes the planet’s climate systems. Alumni use AI to improve healthcare and safety Scott Clark (’08) is pushing the boundaries of artificial intelligence through innovative optimization technologies. After co-founding SigOpt, a startup that helped some of the world’s largest companies fine-tune complex machine learning models and was acquired by Intel for $30M, Clark launched a new venture: Distributional Inc. This latest company focuses on reliability, helping organizations test, calibrate and monitor AI systems — ensuring they behave as expected and can be trusted in real-world use. Jonathan Gallion (’12) wants AI to help modernize human health. As vice president of AI/ ML at OmniScience, he and his team harness large language models and machine learning to modernize clinical trials, accelerate drug discovery, optimize data analysis, and improve decision-making across biotech, medtech, and pharma. His work tackles the explosive growth of healthcare data and ensures AI tools meet real-world requirements — balancing innovation with regulatory and security needs. deep learning → neural networks with many nodes “I am fascinated by the complex relationship we have with our microbiome.” Maude David (pictured above with beneficial probiotic Lactobacillus bulgarius and left with members of her research group) is pioneering a datadriven approach that could transform neuroscience and microbiology alike. IMPACT FALL 2025 11
RkJQdWJsaXNoZXIy MTcxMjMwNg==