Computational design is a commonly used method for generating de novo enzymes. Our project partner, Prof. Zhu Yushan (Tsinghua University, Beijing) has developed the de novo design program PRODA to model the active site of enzymes. However, all de novo design methods still suffers from inherent limitations such as the neglect of long-range electrostatic interactions or the assumption of discrete side chain conformers. As a result, de novo design tools generate a high percentage of false-positive results.
In contrast, molecular dynamics (MD) simulations include long range interactions and have not restriction on the conformational space that is analyzed. As a consequence, MD simulations are more computationally expensive and therefore are the perfect complement to de novo design tools. In the present work, MD simulations are combined with de novo design using PRODA to predict enzymes with desired substrate specificity. Our approach consists of two subsequent steps:
- A comprehensive search for multiple mutants with predicted high catalytic activity by the de novo design tool. As a consequence, a limited number of candidates will be identified, including a large percentage of false positives.
- In a second step, systematic MD simulations of the candidates will be performed to filter out the false positives. It is crucial to determine appropriate geometric descriptors which allow for a reliable distinction between true and false positives.
As a result, a small, highly enriched mutant library will be identified which is sufficiently small to be tested experimentally. The present research project was set up as a feasibility study to identify and validate reliable geometric descriptors.