Title: Assistant Professor, Computer Science, Massachusetts Institute of Technology
Education: PhD, MIT, 2003
Recommended by: Gene Myers
A computer scientist by training, Manolis Kellis has always been fascinated by machines that can adapt to the environment. Before getting into genomic research, Kellis' work centered on reconfigurable robots, artificial intelligence, and pattern recognition. But after becoming interested in the mechanics of the cell, he had a career-altering conversation with a fellow researcher who told him that the cell's algorithm — nature's own program — was already available. “He opened up a file that had pages and pages and pages of ACGT and I was caught staring at my own program, the code that actually makes me work,” says Kellis.
It was also the sheer volume of data that got him intrigued. For his dissertation, Kellis set about mapping the first genome-wide comparison of four complete yeast genomes. He says, “Only after I had published this work and gotten my PhD did I actually hear, 'Hey Manolis, we never told you this, but we all thought it was impossible!'”
Kellis is now focused on elucidating the human genome using model organisms, including yeast. “The power of model organisms is that they allow you to tune in to the particular level of resolution that you're interested in understanding,” he says.
But Kellis and his team are moving beyond what he calls “comparative genomics 101” by discovering specific evolutionary signatures. “By studying these evolutionary signatures of protein coding genes, we can now start to recognize new types of gene events and we can recognize that some segments used a protein with new kinds of signals,” he says. Because many genes don't follow the evolutionary signatures he's finding, Kellis believes that there may be plenty of erroneous genes currently listed in databases. He's already proven that for the yeast genome. After developing gene signatures with the yeast genome, Kellis went back and revised the gene catalogue. Prior to revamping it, researchers had been including 6,300 genes in microarray experiments; after Kellis was through, it turned out only 5,700 genes were real. “This has [had] a tremendous impact on yeast genetics,” he says. “When you probe baker's yeast, everything points to the fact that these are not genes; they are just intergenic regions that were previously mis-annotated.”
Kellis maintains that the only way to continue such refinements is by taking biological discovery in silico. He hopes that soon, instead of relying on traditional genetics to reveal novelties in the genome, computational biology will be at the forefront of discovery.
At some point, Kellis would like to be able to access every step of the evolutionary history of all species. This would include the ability to reconstruct every single ancestral state and to recognize how they transformed over time — as well as where and how every single region is stored in the nucleus. “Basically understanding the 3D state of the nucleus and how DNA is stored and packaged would be tremendous,” he says.
Publications of note
In 2005, Kellis and his collaborators published a paper in Nature called “Systematic discovery of regulatory motifs in human promoters and 3' UTRs by comparison of several mammals.” In it, they showed that it was possible to discover a battery of human motifs at the genome-wide level. They found that roughly 50 percent of 3’ UTR motifs are miRNA-associated with the remainder playing other diverse roles in post-transcriptional regulation.
And the Nobel goes to…
Kellis has grand plans for his Nobel commendation: he aims to win it by reaching an understanding of how to computationally code the geometry for any body plan, including organs and neuro-connections.