Sean Eddy - Washington University - St. Louis: "Biological Sequence Analysis with Probabilistic Models." Talk at the UCSD Bioinformatics Symposium
Sean Eddy had a very interesting talk, obviously departing from his normal presentation to one that addressed the history of his career in science. It was very interesting to hear from a pioneer in the field of probabilistic models in bioinformatics.
He got his start looking for similar intron/RNAs. These are aligned using paired bases. What he needed was a way to simultaneously align and score (structurally) a sequence to a template. Originally he was working on a purely wet-lab RNA project and was scooped so went with what he was moonlighting as computational biology. Luckily, his postdoc committee didn't care that he completely dropped what he originally started with. Talking to people in other departments helped him crystallize his ideas. Talked to speech people at Cambrige, England to get good feedback.
After thinking about the need for a better RNA alignment, he got handed the Karplus, Haussler SAM paper about aligning sequences with HMMs. After reading the paper he was inspired to apply this to his RNA problem. It started with a complete reimplementation of SAM, in a package that he named hmmer. He said that this was a good lesson, if you really want to understand something just go and reimplement it. Working out the bugs and nuances will teach you more than just thinking about the problem.
It was a very interesting talk and although I had never met Sean Eddy before, I found him to be a very nice guy. I never would have guessed from his standoffish website, which has undoubtedly put off many a potential student.