1. Embedding Computation in Studies of Protein Structure and Function Chris Bailey-Kellogg & lab & collaborators Shobha Xiaoduan Wei John Chris Fei

2. What does it look like? Generate models Optimize set of experimental probes Evaluate models wrt data Characterize differences

3. What does it look like? Represent entire sol’n space Prune parts inconsistent w/data Evaluate uncertainty in remainder

4. What does it do (how, why)? K V A V V L Q W I L K W G F A L Q C P K S C A L V V L S I A D L A V V L R G I T T C I I L L H G A M E M L H I L P V L S E L L T V L P N V D Q K A T V L L N I F R I F Y P L I S K Y C M I L A L M T V E C T A V V L M A I K Analyze evolutionary record Infer graphical model Analyze, predict, classify

5. What does it do (how, why)? Represent NMR data as graph Represent structure as graph Use isomorphisms to localize interactions, dynamics

6. How can we change it? Recombination: mix-and-match fragments from related proteins Evaluate breakpoint locations wrt constraints, diversity Optimize by dynamic programming

7. How can we change it? Optimize robotic assembly of hybrid library

8. Embedded Computation Minimize experimental complexity Be robust to sparsity and noise Maximize information gain Employ appropriate representations Search efficiently over possibilities Combine sources of information Quantify uncertainty Change approach to experiment and analysis: assumptions, experiments, and overall process