Studies in Computational Biochemistry: Applications to Computer Aided Drug Discovery and Protein Tertiary Structure Prediction
Year: 2019
Type: article
Abstract: Computational biochemistry is an ever-growing field that involves the use of computerbased technology to study biochemical problems.This dissertation focuses on two applications of computational biochemistry: computer aided drug discovery and protein structure prediction.The first portion of this work, comprising Chapter 2, details a study performed to identify novel compounds that increase the calcium sensitivity of cardiac troponin C, a protein involved in heart muscle contraction.This was accomplished using a combination of virtual screening and a relaxed complex scheme.A set of highly predictive cardiac troponin C structures were identified using a receiver operator characteristic (ROC) analysis and a set of known binding compounds.Using these structures as the receptors for structure-based drug discovery, a virtual screen was performed on the National Cancer Institute's Developmental Therapeutic Program database.The screen followed by experimental verification yielded two novel compounds, NSC600285 and NSC611817, that demonstrated increased calcium sensitivity.The second portion of this work, detailed in Chapters 3 and 4, focuses on the development of the methodology to incorporate sparse covalent labeling mass spectrometry data into Rosetta's ab initio structure prediction protocol.In Chapter 3, covalent labeling data in the form of hydroxyl radical footprinting was used to develop a new centroid based score term iii for the Rosetta energy function.This new score term, called hrf_ms_labeling, was used to rescore models from a benchmark set of four soluble proteins with available labeling data.Upon rescoring, the accuracy of the predicted models improved as compared to Rosetta prediction alone.In the case of two of the proteins, atomic resolution models were identified.As a continuation of the work with hydroxyl radical footprinting, the study presented in Chapter 4 sought to expand upon the use of covalent labeling data with Rosetta.Two new score terms, covalent_labeling_cen and covalent_labeling_fa, were added to Rosetta that use a "cone"-based neighbor count as a metric of solvent exposure.These new score terms were used to rescore structures predicted by Rosetta, much like the HRF based score term.The terms were also incorporated into the AbinitioRelax protocol and used to generate protein structure predictions that were guided by covalent labeling data.Additionally, rather than focus on a single labeling reagent, a variety of covalent labeling techniques were computationally analyzed.Using twenty proteins with known experimental structures pulled from the Protein Data Bank as a test set, an optimal set of residue types for labeling were identified.The tolerance of the score terms to false negative and false positive data points was also investigated.Combining all of this information resulted in an average accuracy improvement of 3.9 Å RMSD upon rescoring the test protein set.Generating models guided by the covalent labeling scores also resulted in improvement.sister Alexis, my mother-in-law and father-in-law Debbie and Steve, and my sister-in-law and brother-in-law Carly and Stephen for being with me for every step of this adventure.I am endlessly appreciative for all the love and support they have given me throughout this process.I would like to thank
Author: Aprahamian, Melanie Lorraine
Institutions: Pennsylvania Western University, The Ohio State University
Language: English
Subfield: Molecular Biology
Domain: Life Sciences
Open Access status: green
OpenAlex ID: W7071588705
