Record:   Prev Next
作者 Van Voorst, Jeffrey Ryan
書名 Surface matching and chemical scoring to detect unrelated proteins binding similar small molecules
國際標準書號 9781124775562
book jacket
說明 205 p
附註 Source: Dissertation Abstracts International, Volume: 72-10, Section: B, page: 6126
Advisers: Leslie A. Kuhn; Yiying Tong
Thesis (Ph.D.)--Michigan State University, 2011
How can one deduce if two clefts or pockets in different protein structures bind the same small molecule if there is no significant sequence or structural similarity between the proteins? Human pattern recognition, based on extensive structural biology or ligand design experience, is the best choice when the number of sites is small. However, to be able to scale to the thousands of structures in structural databases requires implementing that experience as computational method. The primary advantage of such a computational tool is to be able to focus human expertise on a much smaller set of enriched binding sites
Although a number of tools have been developed for this purpose by many groups [53, 63, 89, 91, 94], to our knowledge, a basic hypothesis remains untested: two proteins that bind the same small molecule have binding sites with similar chemical and shape features, even when the proteins do not share significant sequence or structural similarity. A computational method to compare protein small molecule binding sites based on surface and chemical complementarity is proposed and implemented as a software package named SimSite3D. This method is protein structure based, does not rely on explicit protein sequence or main chain similarities, and does not require the alignment of atomic centers. It has been engineered to provide a detailed search of one fragment site versus a dataset of ∼ 13,000 full ligand sites in 2-4 hours (on one processor core)
Several contributions are presented in this dissertation. First, several examples are presented where SimSite3D is able to find significant matches between binding sites that have similar ligand fragments bound but are unrelated in sequence or structure. Second, including the complementarity of binding site molecular surfaces helps to distinguish between sites that share a similar chemical motif, but do not necessarily bind the same molecule. Third, a number of clear examples are provided to illustrate the challenges in comparing binding sites which should be addressed in order for a binding site comparision method to gain widespread acceptance similar to that enjoyed by BLAST [3, 4]. Finally, an optimization method for addressing protein (and small molecule) flexibility in the context of binding site comparisons is presented, prototyped, and tested
Throughout the work, computational models were chosen to strike a delicate balance between achieving sufficient accuracy of alignments, discriminating between accurate and poor alignments, and discriminating between similar and dissimilar sites. Each of these criteria is important. Due to the nature of the binding site comparison problem, each criterion presents a separate challenge and may require compromises to balance performance to achieve acceptable performance in all three categories
At the present, the problem of addressing flexibility when comparing binding site surfaces has not been presented or published by any other research group. In fact, the problem of modeling flexibility to determine correspondences between binding sites is an untouched problem of great importance. Therefore, the final goal of this dissertation is to prototype and evaluate a method that uses inverse kinematics and gradient based optimization to optimize a given objective function subject to allowed protein motions encoded as stereochemical constraints. In particular, we seek to simultaenously maximize the surface and chemical complementarity of two closely aligned sites subject to directed changes in side chain dihedral angles
School code: 0128
Host Item Dissertation Abstracts International 72-10B
主題 Chemistry, Biochemistry
Biology, Bioinformatics
Biophysics, General
Computer Science
0487
0715
0786
0984
Alt Author Michigan State University. Computer Science
Record:   Prev Next