Title: Assistant Professor of Medicine and Pediatrics, Stanford School of Medicine
Education: PhD, Harvard and MIT, 2004; MD, Brown University, 1995
Recommended by: Alan Krensky
Atul Butte and his team are tasked with developing methods to integrate disparate disease-related data sets garnered from more than 30 types of high-throughput measurement technologies currently available. Butte sees his mission as pushing nosology, the disease classification system first developed in the 18th century, into the 21st century via a genomics approach. He believes that the nosology system still in use today is long overdue for a genomics overhaul.
That's why Butte and his team set out to develop a genomic classification scheme for medicine called “genomed”: genomic nosology for medicine. “Why don't we think about classifying diseases based on genomics?” Butte says. “Instead we use symptoms: all the cancers go together, or things that cause headaches go together, but it has nothing to do with modern science.”
His team's aim is to develop an automated network capable of drawing inferences across the breadth, depth, and width of molecular biology data, including entire sets of transcripts, proteins, and genes. Butte, who is also a physician, is primarily concerned with bringing his integrative methods to bear on what he sees as the coming health crisis: obesity and type 2 diabetes mellitus. Currently, his lab works in conjunction with the Joslin Diabetes Center on integrating multiple types of genome-scale data across experiments and phenotypes to identify genes involved in these diseases.
To begin the momentous task of reclassifying diseases based on genomic data, Butte and his lab will collect disease-related data sets from international repositories such as the Gene Expression Omnibus, which has more than 100,000 microarray samples. “What we did was literally download every single experiment that's out there, we figured out every single piece of experimental details that they've done: Did they look at aging? Did they look at injury? Did they look at leukemia? We were trying to figure out the experimental context using microarray annotations and figure out the gene measurements simultaneously,” Butte says. He and his team have written software that correlates the experimental descriptions and codes them according to a structured vocabulary built by the National Library of Medicine.
Looking ahead
Butte would like to see bioinformatics eventually move away from a service mode to exploring the seemingly endless connections among disease data. But the hurdle practitioners face is finding a lab collaborator to validate those findings. “I see too many bioinformaticians just working on the next great method to analyze microarray data, or the next best way to analyze mass spec data from proteomics, but instead they could ask a question like 'What's in common between this disease and this disease?'” says Butte. “You're not beholden to any one biologist, and when you have an idea about a disease, you go to a biologist and say 'You know, I've got 100,000 microarrays and I feel that these genes are in common and this is what it means, wouldn't you like to help validate me?' I bet no one would say no to that. It puts you in the driver's seat.”
Publications of note
This year, Butte and Isaac Kohane published a paper entitled “Creation and implications of a phenome-genome network” in Nature Biotechnology. In it, they used what Butte calls “traditional informatics tools,” such as ontologies and structured vocabularies, to connect every part of every disease and environmental factor to the genes that go along with those factors.
And the Nobel goes to…
Butte would like to accept his Nobel for “finding a cure for type 2 diabetes mellitus by applying methods we created for use in integrative biology.”
— MD