[denzil_correa]

> We're working on leveraging sparse experimental data for protein structure prediction. To do that, we end up using algorithms and ideas from different various CS fields, from machine learning, to robotics, to distributed systems. Other people are working on exciting fields like computation protein design, studying drug protein interaction in silico..

Are there particular methods you use to deal with little and sparse data?

[resiros]

We're basically merging the sparse experimental data we get with other priors we have (the energy landscape, residue-residue contacts predicted from evolutionary data) in a Expectation Maximization kind of algorithm, where each step you get better predictions (in the sense of they satisfy the experimental data while agreeing with the priors from the problem (low energy, nice fold..).