Staphyloccocus aureus is an opportunisitic pathogen that causes soft skin and tissue infections (SSTI) such as endocarditis, osteomyelitis and meningitis. In recent years, the re-emergence of antibiotic-resistant S. aureus such as MRSA presents a formidable challenge for infection management worldwide. Amidst this global epidemic of antimicrobial resistance, several research efforts have turned their focus towards exploiting the cell-wall biosynthesis pathway for novel anti-bacterial targets. Recently, the lipoteichoic acid (LTA) biosynthesis pathway has emerged as a potential anti-bacterial target. LTA is an anionic polymer found on the cell envelope of Gram-positive bacteria. It comprises of repeating units of glycerol-phosphate (GroP) and is important for bacterial cell physiology and virulence. For example, it is critically involved in regulating ion homeostasis, cell division, host colonization and immune system invasion. Several reports showed that bacteria lacking LTA are unable to grow. At the same time, they suffer from severe cell division defects and also exhibit aberrant cell morphologies. The key protein involved in the LTA biosynthesis pathway is the Lipoteichoic acid synthase (LtaS). LtaS is located on the cell membrane of Gram-positive bacteria and can be divided into two parts: a transmembrane domain and an extra-cellular domain responsible for its enzymatic activity (annotated eLtaS). Given that LtaS is important for bacterial survival and there are no known eLtaS homologues in eukaryotic cells, this protein is an attractive antibacterial target. In 2013, a small molecule eLtaS inhibitor (termed 1771) was discovered. Although 1771 was able to deplete LTA production, the binding mechanism of 1771 to eLtaS remains unknown. Additionally, 1771 could only prolong the survival of infected mice temporarily because of its in vivo instability. Therefore, the need for finding more potent and metabolically stable inhibitors of eLtaS still remains. Computational-aided drug design (CADD) is a cost-effective and useful approach that has been widely integrated into the drug discovery process. The protein eLtaS lends itself to be a good target for CADD since its crystal structure and a known inhibitor (with limited structure-activity data) is available. In this work, I have targeted eLtaS using CADD methodology followed by prospective validation using various biophysical, biochemical and microbiological assays. My project can broadly be sub-divided into three phases: (a) identification of small molecule binding “hot spots”, (b) optimization of existing inhibitor and (c) discovery of new hits. Through a systematic use of different computational approaches, I modelled a plausible 1771-bound eLtaS complex and used the structural insights to generate new inhibitors against eLtaS. To this end, I discovered EN-19, which is a more potent inhibitor of eLtaS. Additionally, by targeting transient cryptic pockets predicted by Molecular Dynamic simulations, I have discovered a new inhibitor chemotype that seems to exhibit a different mode of action against eLtaS. Taken together, my work presents a computational platform for future drug design against eLtaS and reinforces the notion that targeting eLtaS is a viable strategy to combat Gram-positive infections.

Machine learning (ML) has enjoyed great success in chemistry and drug design, from designing synthetic pathways to drug screening, to biomolecular property predictions, etc.. However, ML model's generalizability and robustness require high-quality training data, which is often difficult to obtain, especially when the training data is acquired from experimental measurements. While one can always discard all data associated with noisy and/or missing values, this often results in discarding invaluable data. This thesis presents and applies mathematical techniques to solve this problem, and applies them to problems in molecular medicinal chemistry. In chapter 1, we indicate that the missing-data problem can be expressed as a matrix completion problem, and we point out how frequently matrix completion problems arise in (bio)chemical problems. Next, we use matrix completion to impute the missing values in protein-NMR data, and use this as a stepping-stone for understanding protein allostery in Chapter 2. This chapter also used several other techniques from statistical data analysis and machine learning, including denoising (from robust principal component analysis), latent feature identification from singular-value decomposition, and residue clustering by a Gaussian mixture model. In chapter 3, Δ-learning was used to predict the free energies of hydration (Δ𝐺). The aim of this study is to correct estimated hydration energies from low-level quantum chemistry calculations using continuum solvation models without significant additional computation. Extensive feature engineering, with 8 different regression algorithms and with Gaussian process regression (38 different kernels) were used to construct the predictive models. The optimal model gives us MAE of 0.6249 kcal/mol and RMSE of 1.0164 kcal/mol. Chapter 4 provides an open-source computational tool Procrustes to find the maximum similarities between metrics. Some examples are also given to show how to use Procrustes for chemical and biological problems. Finally, in Chapters 5 and 6, a database for permeability of the blood-brain barrier (BBB) was curated, and combined with resampling strategies to form predictive models. The resulting models have promising performance and are released along with a computational tool B3clf for its evaluation. / Thesis / Doctor of Science (PhD)