João M. Martins:
Computational Studies of Drug Resistance. Investigating Binding Affinities and Structural Ensembles of Proteins

Date: 27-04-2017    Supervisor: Kresten Lindorff-Larsen



Drug resistance has been an increasing problem in patient treatment and drug development. Starting in the last century and becoming a major worry in the medical and scienti c communities in the early part of the current millennium, major research must be performed to address the issues of viral and bacterial resistance to common-use inhibitors, in such cases as multiple targeted proteins in the human immunode ciency virus infection and penicillinresistant Staphylococcus aureus. Thus, understanding the evolutionary pressures by which these arise and predicting future possible resistance mutations is of the utmost importance in developing better and less resistance-inducing drugs. A drug's in uence can be characterized in many diff erent ways, however, and the approaches I take in this work re ect those same different in uences. This is what I try to achieve in this work, through seemingly unrelated approaches that come together in the study of drug's and their in uence on proteins and vice-versa.

In part I, I aim to understand through combined theoretical ensemble analysis and free energy calculations the e ects mutations have over the binding anity and function of the M2 proton channel. This research on mutants' e ect is performed in coordination with experimental work performed by colleagues, with the experimental mutant development directing this theoretical work in terms of mutants tested and drugs used in order for a comparison to be made. For this reason, the manuscripts arising from this work are included in this part's Supporting Information. While some correlation between experimental and theoretical results can be observed through the study of the proteins' ensembles and drugs' behavior in simulation, I also observe the limitations of the methods employed in distinguishing anities between low-anity mutants and propose some alternative approaches which could have been employed towards this goal.

Work in part II focuses on ensembles and prediction of experimental measurements through the use of ensemble information. This is done by implementing and expanding an existing rotamer library approach method for double electron-electron resonance prediction. The comprehensiveness and speed of this implementation lead to the development of novel implementations being performed for other probe-dependent experimental methods in which ensemble information is important, paramagnetic relaxation enhancement and Forster resonance energy transfer calculations. As is the case for theoretical prediction methods, a benchmark of the prediction must be performed, which yielded good results in accordance with earlier experimental measurements. With these tools in hand, ensemble perturbation by drug binding can be studied, as an example is discussed in the Conclusions and Further Research section.

With all this in place, I believe this work has furthered the understanding of an important drug resistant system such as the M2 proton channel. To this date and to the best of my knowledge, no such structural work has been done with as many drugs or mutants tested, with important insights being found about the structure of the channel upon mutation and the positioning of drugs inside the channel. I have also put in place new methods by which a simulation can be compared to experimental results, useful in simulation validation, ensemble perturbation and experimental planning. Through this work, I expanded the usefulness of an well established prediction method, as well as generated new state-of-the-art implementations of not yet performed ensemble back-calculations, opening the door for further uses and expansion in the study of protein ensembles in solution.