Content area
Developing new drug is a long and costly process: it takes on average more than ten years and costs over $2 billion to bring a drug to market. Drug development is a multi-disciplinary initiative involving not only chemistry and biology, but also pharmacology, toxicology, biophysics, computer and data science, to name a few. The development of artificial intelligence has nowadays lead to a revolution in many areas, and pharmaceutical industry is not an exception. The use of computational techniques to aid drug discovery started decades ago and participated in many aspects of drug discovery, such as proposing novel drugs by virtually screening molecular libraries, as well as predicting toxicity, solubility and stability. Scientists keep developing efficient and accurate computational tools and apply them in drug discovery projects as a regular routine in order to shorten the development time and reduce the cost of production. In this thesis, my research on both the development of computational chemistry methods and applications to a drug discovery project is described.
My research is related to computer-aided structure-based drug design. This type of approaches requires detailed 3D structural information of a drug target like nucleic acids, and predicts the potentially active molecules through a virtual evaluation of binding between ligands and targets. This can be done for example by computing the binding energies. One time-efficient way is to model atoms as spheres and compute the interactions based on classical mechanics with predefined functions and parameters, referred to as molecular mechanics (MM). The first part of the thesis aims to explain how a series of platinum (II) complexes bind to a set of DNA G-quadruplex structures through a MM-based simulation. This part of the study reveals a common limitation of using MM in the modeling of novel molecules: although parameters for common molecular fragments have been developed, parameters describing structurally new molecules are always not available. Additional effort is required from the user to develop such parameters, which requires time and computing resources. In the rest of the thesis, this problem was addressed by systematically developing MM parameters based on quantifying chemical principles. Sterics, hyperconjugation, inductive and resonance effects have been qualitatively utilized for decades to rationalize experimental observations. As a proof-of-concept, we have demonstrated that hyperconjugation is a major contributor to the conformational energy and can be computed from atomic properties on-the-fly. A method called H-TEQ (Hyperconjugation for Torsional Energy Quantification) was developed. Next, the effect of hyperconjugation from lone pairs was studied and an updated version of H-TEQ, able to compute hyperconjugation for all types of saturated molecules, was developed. Lastly, the effect of conjugation was quantitatively studied, with the aim of developing a method to also cover conjugated molecules.