Zhang, Xia2014-04-232014-04-232014-02https://hdl.handle.net/11299/163034University of Minnesota Ph.D. dissertation. February 2014. Major: Medicinal Chemistry. Advisor: Elizabeth A. Amin. 1 computer file (PDF); xii, 180 pages.Rtt109 is a fungal-specific histone acetyltransferase that catalyzes histone H3 lysine 56 acetylation and is a promising antifungal drug target. To identify novel Rtt109 inhibitors as potential drug scaffolds, we employed in vitro high throughput screening (HTS) and various computer-assisted strategies, including molecular dynamics, docking, three-dimensional quantitative structure-activity relationship (3D-QSAR) analysis, pharmacophore modeling, and Support Vector Machine (SVM) mining. An initial experimental screening of 82,861 compounds (HTS1) yielded hits with activity ranging from 0.49 - 17.5 µM against Rtt109. The molecular dynamics simulation of Rtt109 suggested that the histone lysine tunnel, a potential inhibitor binding site, was not flexible and thus the use of a rigid protein structure of Rtt109 was appropriate for docking studies. From a virtual screen using Surflex-Dock, we have identified 878 additional compounds as potential hits, with predicted Kd values of 0.1 nm or lower. Based on preliminary experimental data from HTS1, validated pharmacophore maps were developed and used to pinpoint potential Rtt109 ligand-receptor interactions. 3D-QSAR CoMFA and CoMSIA models that were also derived from the hit series generated in the initial experimental HTS display high self-consistency (r<super>2</super> = 0.985 [CoMFA] and r<super>2</super> = 0.976 [CoMSIA]) and robust internal predictivity (r_cv<super>2</super> = 0.754 [CoMFA] and r_cv<super>2</super> = 0.654 [CoMSIA]). Importantly, key features identified in both the pharmacophore hypotheses and the 3D-QSAR models agreed well with each other and with experimentally defined structural features in the Rtt109 lysine-binding tunnel. In addition, our optimized SVM models demonstrated high predictive power against the external test sets for Rtt109 with accuracy of 91.1%. We also identified novel features with significant differentiating ability to separate Rtt109 inhibitors from non-inhibitors.en-USThesis or Dissertation