AMS Invited Address speaker Bonnie Berger gave a captivating talk on Biomedical Data Sharing. There is a recurring issue in Computational Biology: genomic data is growing exponentially faster than computing power and data storage. If we want researchers to have access to the data, compressing data, sharing it, and decompressing it is not a viable solution due to the fact that decompressing is time consuming and requires a lot of storage. Bonnie’s solution: Compress the data and operate on the compress data. How?
Bonnie’s research team provided an algorithm that plots data points in a high-dimensional Euclidean space and covers the data points with spheres.
Then, by only using one point per sphere, the algorithm searches through the spheres, decides which are of interest to the particular problem of study and then goes back to these spheres and does a more thorough search in a region slightly larger than these spheres. The result? An algorithm that is much faster than the available methods and that recovers more than 99% of the results that slower algorithms recover. This work has been cited thousands of times.
Bonnie then went to describe other problems for which they provided similar algorithms to reduce run time while providing great accuracy. One that seemed really interesting was to combine and cluster cell data that was obtained via multiple experiments by different research groups.
For more information, visit Bonnie Berger’s website: http://people.csail.mit.edu/bab/computing.html