An immediate reduction in global CO2 emissions could be accomplished by replacing coal- or oil-based energy sources with purified natural gas. The most important process involved in natural gas purification is the separation of CO2 from CH4, where Pressure Swing Adsorption (PSA) technology on porous materials has emerged as a less energy demanding technology. Among porous materials which are used or could potentially be used in PSA, Metal Organic Frameworks (MOFs) have attracted particular interest owing to their record-breaking surface areas, high-porosity, and high tunability. However, the discovery of optimal MOFs for use in adsorption-based CO2 separation processes is remarkably challenging, as millions of MOFs can potentially be constructed from virtually limitless combinations of inorganic and organic secondary building units. To overcome this combinatorial problem, this thesis aims to (1) identify important design features of MOFs for CO2/CH4 separation through the investigation of currently existing MOFs as well as the high throughput computational screening of a large database of MOFs, and to (2) develop efficient computational tools for aiding the discovery of new MOF materials. To validate the computational methods and models used in this thesis, the first work of this thesis presents the computational modeling of CO2 adsorption on an experimental CuBDPMe MOF using grand canonical Monte Carlo simulations and density functional theory. The simulated CO2 adsorption isotherms are in good agreement with experiment, which confirms the accuracy of the models used in our simulations throughout this thesis. The second work of this thesis investigates the performance of an experimental MIL-47 MOF and its seven functionalized derivatives in the context of natural gas purification, and compares their performance with that of other well-known MOFs and commercially used adsorbents. The computational results show that introducing polar non-bulky functional groups on MIL-47 leads to an enhancement in its performance, and the comparison suggests that MIL-47-NO2 could be a possible candidate as a solid sorbent for natural gas purification. This study is followed by the compactional study of water effects on natural gas purification using MOFs, as traces of water is present in natural gas under pipeline specifications. From the study, it is found that water has a marginal effect on natural gas purification in hydrophobic MOFs under pipeline specifications. Following the aforementioned studies, a database of 324,500 hypothetical MOFs is screened for their performance in natural gas purification using the general protocol defined in this thesis. From the study, we identify 'hit' materials for targeted synthesis, and investigate the structure-property relationships with the intent of finding important MOF design features relevant to natural gas purification. We show that layered sheets consisting of poly-aromatic molecules separated by a perpendicular distance of roughly 7 Å are an important structural-chemical feature that leads to strong adsorption of CO2. Following the screening study, we develop efficient computational tools for the recognition of high-preforming MOFs for methane purification using Machine Learning techniques. A training set of 32,500 MOF structures was used to calibrate support vector machines (SVMs) classifiers that incorporate simple geometrical features including pore size, void fraction and surface area. The SVM machine learning classifiers can be used as a filtering tool when screening large databases. The SVM classifiers were tested on ~290,000 MOFs that were not part of the training set and could correctly identify up to 70% of high-performing MOFs while only flagging a fraction of the MOFs for more rigorous screening. As a complement to this study, we present ML classifier models for CO2/CH4 separation parameters that incorporate separately the Voronoi hologram and AP-RDF descriptors, and we compare their performance with the classifiers composed of simple geometrical descriptors. From the comparison, it is found that including AP-RDF and Voronoi hologram descriptors into the classifiers improves the performance of classifiers by 20% in capturing high-performing MOFs. Finally, from the screening data, we develop a novel chemiformatics tool, MOFFinder, for aiding in the discovery of new MOFs for CO2 scrubbing from natural gas. It has a user-friendly graphical interface to promote easy exploration of over 300,000 hypothetical MOFs. It enables synthetic chemists to find MOFs of interest by searching the database for Secondary Building Units (SBUs), geometric features, functional groups and adsorption properties. MOFFinder provides, for the first time the substructure/similarity query of porous materials for users and is publicly available on titan.chem.uottawa.ca/moffinger.

No description available.

The rapid pace of scientific advances is enabling a greater understanding of diseases at the molecular level. In turn, the process for researching and developing new medicines is growing in difficulty, costs, and length as a result of the scientific, technical, and regulatory challenges related to the development process. In light of these challenges, drug repositioning, the utilization of known drugs for a new medical indication, has emerged as an increasingly important strategy for the new drug discovery. Availability of prior knowledge regarding safety, efficacy and the appropriate administration route significantly reduces the development costs and cuts down the development time resulting in less effort to successfully bring a repositioned drug to market. In another aspect, a protein’s shape is closely linked with its function; thereby, the ability to predict this structure unlocks a greater understanding of what it does and how it works. Nowadays, more than 10,000 biologically relevant protein structures are yearly released and available to the scientific community. A number suspected to triple over the following years due to the recent breakthroughs in structure prediction techniques. This work introduces a novel structure-based drug repositioning approach, exploiting the similarities of drugs’ binding mode via identification and virtual screening of interaction patterns. Such patterns are uncovered with the use of PLIP, an automated tool for the in silico detection of non-covalent interactions defining the binding mode between drugs and their protein targets. Besides, the approach has been applied to a series of case studies with tangible results: the uncovering of an antimalarial drug as potential chemoresistance treatment, the explained binding mode of ibrutinib to the target VEGR2 as potential B-cells deactivator in autoimmune diseases, and three over the counter drugs with a proved anti-trypanocidal activity as treatments for Chagas disease. Overall the structure-based approach with interaction patterns proved to be a suitable framework for identifying novel repositioning candidates. The uncovered candidates were structurally unrelated to the currently available treatments, and experimental assays successfully demonstrated their inhibitory activity on the protein targets of interest. Furthermore, the approach represents a promising option for the 'in high demand' diseases and all rare and neglected diseases for which no reliable treatment has yet been found and for which the pharmaceutical industry makes only a little investment.