The ability to use sequence alone to quickly and confidently predict free energies of association for interacting proteins has significant implications for biomedical and biochemical research. For example, consider the Kazal family protein inhibitor OMTKY3 (turkey ovomucoid third domain) and enzyme SGPB (Streptomyces griseus proteinase B). The above figures illustrate the OMTKY3 sequence and an x-ray structure of the bound complex. An algorithm for predicting the effect on binding of proposed amino acid substitutions in OMTKY3 could support the design of a modified inhibitor that is tighter-binding, more specific to SGPB (relative to, say, chymotrypsin), and so forth. Similarly, applying this algorithm over known sequences in the Kazal family could support the study of evolutionary hypotheses regarding selection for binding specificity vs. binding strength.

We are developing a combined computational-experimental approach to recognize and use additivity in predicting and analyzing distributions of free energies of association for entire protein families. For example, OMTKY3 has 10 variable contact residues (in blue in the sequence above) which highly impact the binding strength; by varying them, we can evaluate the set of all possible inhibitors. Due to the exponential number of possible sequences, efficient algorithms are required to characterize the distribution of binding strength over such a set. Given such a characterization, additional algorithms are required to optimize specificity of binding against different enzymes, analyze differences in distributions (all possible, natural sets, different enzymes, etc.) in order to reverse-engineer nature's optimization, and so forth.

We have made available a simple web-based analysis tool for substitutions to OMTKY3 based on some published data from the Laskowski lab.

• Jiangtian Li
• Olga Vitek

• Michael Laskowski (chemistry, Purdue)
  Postdocs Ying Gao and ZhengPing Yi