Adaptation and Use of Four-Body Statistical Potential to Examine Thermodynamic Properties of Proteins A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy at George Mason University By Gregory M. Reck Bachelor of Science University of Cincinnati, 1969 Director: Iosif I. Vaisman, Associate Professor Program in Bioinformatics and Computational Biology Fall Semester 2006 George Mason University Fairfax, VA UMI Number: 3240838 Copyright 2006 by Reck, Gregory M. All rights reserved.
INFORMATION TO USERS The quality of this reproduction is dependent upon the quality of the copy submitted. Broken or indistinct print, colored or poor quality illustrations and photographs, print bleed-through, substandard margins, and improper alignment can adversely affect reproduction. In the unlikely event that the author did not send a complete manuscript and there are missing pages, these will be noted. Also, if unauthorized copyright material had to be removed, a note will indicate the deletion.
® UMI UMI Microform 3240838 Copyright 2007 by ProQuest Information and Learning Company. All rights reserved. This microform edition is protected against unauthorized copying under Title 17, United States Code. ProQuest Information and Learning Company 300 North Zeeb Road P.
Box 1346 Ann Arbor, MI 48106-1346 ii Copyright 2006 Gregory M. Reck All Rights Reserved ADAPTATION AND USE OF FOUR-BODY STATISTICAL POTENTIAL TO EXAMINE THERMODYNAMIC PROPERTIES OF PROTEINS by Gregory M. Reck A Dissertation Submitted to the Graduate Faculty of George Mason University in Partial Fulfillment of The Requirements for the Degree of Doctor of Philosophy Bioinformatics Committee: “2⁄2. Director Program Director Assk\st Department Chairperson Zp a.
yr Chm LÍ Aa Dean, College of Science Date: | Y [ Ols Fall Semester 2006 George Mason University Fairfax, VA iii DEDICATION This dissertation is dedicated to my loving wife, Lelia, to Chelsea, the best daughter in the world and to the memory of my parents, Marie and Ronald Reck. ACKNOWLEDGEMENTS The opportunity to pursue a passion may not occur for everyone, but I have been blessed with that chance and it was due primarily to my wife, Lelia. Her encouragement and support from the outset have been unwavering. She set the example for a career path in science, challenged me to follow, and has truly been a source of inspiration.
I was indeed fortunate to have the opportunity to work with my advisor, Dr. Vaisman, a talented educator, a gifted chess player and a true gentleman. He has provided just the right mix of guidance, review and advice to keep things moving forward. And I particularly enjoyed our meetings at La Madeleine.
My sincere appreciation to my committee members, Dr. Jamison and Dr. Born for taking the time to support my efforts. I took my first class at GMU from Dr.
Willett and his insights and enthusiasm convinced me that I should press on. Jamison helped me get underway as my first advisor, and stayed with me even after his relocation to Northern Kentucky University. Born is genuinely intent on finding linkages between bioinformatics and his work with enzymes. I also want to thank Majid Masso, Ewy Mathe, Todd Taylor and Andrew Carr for the opportunity to exchange ideas and share experiences.
I appreciate their acceptance of someone from a different generation. A special thanks to Glenda Wilson, the magician who really makes things happen for the Bioinformatics students. And also to Chris Ryan for keeping the computers well- behaved and under control. Finally, I want to acknowledge the support of NASA management in providing me with the opportunity to get started in bioinformatics and recognizing that there may be alternatives to conventional wisdom in space exploration.
TABLE OF CONTENTS Page 9ò nh.1 Computational Approaches to Folding and Stability Analysis.3 Specific Aims and Dissertation FOFTât. G111 g1 ng gen 10 2 A Novel Approach to Protein-water Interaction Characteristics Using Computational GeOmMetry.2 Computational Hydration of the Protein Set.4 DT Simplex Face Match Residue Classification.5 Water Coordination Number RaiO.6 DT Water Group Parameter and Residue Classification .7 Accessible Surface AT€A. HH ch HT HH 0 0 0001 0 97 27 2.9 Circular Variance Tan.4 Results and DisCussiOn.1 Examination of the Protein HydratiOns.- - HH ng se 29 2.2 Comparison of Tessellation Classification Methods. Relationship of Simplex Face Matching Method to Location Parameters .4 Relationship of Water Group Method to Location Parameters.5 Application of the Residue Classification Methods to a Specific Protein.6 Examination of the Water Group Parameter and the Water CNR.7 Comparison of Water Group Parameter with Hydrophobicity Scales.
Q0 Họ TH 0004 06 80 04 55 3 Nearest-neighbor Contact Potentials Derived From Delaunay Tessellation of Hydrated Protein. - HH HH nh TT 0 60 ch.4 Results and DiSCUSSIOTI.1 Characteristics of the Potential Functions. do - co c1 ng gu 76 3.2 Decoy Discrimination Using the Tessellation Potential Functions. Comparison of Tessellation Results With Reported Models for Decoy BC osii 11.
93 4 Use of Statistical Potentials Derived From Delaunay Tessellation to Characterize Changes in Protein Stability Due to Single Point Mutations.- - -Q c HH Họ th 90096 101 4. Derivation of Statistical PotentiaÌS.-- --- ng ng vấn 104 4.4 Application of Statistical Potential Functions to Target Proteins .5 Machine Learning ToOÌS. cv HH ng nọ ng 00096 113 `.1 Application of Tessellation Potential to the Study Proteins.2 Comparison of Statistical Potential Strategies (CA, WG and SP).3, Correlations With Specific Residue Types. Use of Machine Learning Tools to Identify Stability Content in Mutant Residual PTrofiles.5 Examination of Transthyretin Mutant Residual Profiles for Amyloid Signal Using Machine Learning ToOÌS.
158 5 Future Dir€CtÏOnS.1 Residual Profile Searches.2 Exploration of Amyloid Mutants with Machine Learning Tools. Comparison of Water Group Parameter with Hydrophobicity Scales.G - Ă SH HH He 163 vii LIST OF TABLES Table Page Table 2.1 Comparison of the proximity of Dowser placements with the nearest crystal HOH and SOLVATE sites for 9 selected proteins.2 Number of residues classified by the simplex face match and the water group methods in specific protein ]OCAtÏOTIS. - - cọ HH ng gà 41 Table 3.1 Results of the decoy discrimination ranking of the decoy sets using various Statistical potential functions derived from both C-alpha and center of mass residue LOCATIONS.2 Results of the decoy discrimination ranking of the decoy sets using each of the three potential strategies with added hydrogen atoms. sen seesee 88 Table 3.3 Comparison of decoy discrimination ranking with reported models .1 Comparison of potentials for study proteins.2 Ability of three potential functions to correlate mutant stability changes in barnase and staphylococcal nuclease.3 Ability of three potential functions to correlate mutant stability changes in T4 LYSOZYME.4 Correlations of stability with potential for specific residue types in Staphylococcal nucÌease.-- - Ác ng HT ng ng ngà 130 Table 4.5 Performance of the Random Forests ML tool for predicting amyloid behavior for mutants of human transthyretin.- - HH ng 9031955 153 viii LIST OF FIGURES Figure Page Figure 2.1 Illustration of a tessellated hydrated protein.2 Effect of hydration on the non-redundant edge lengths of simplices.3 Illustration in two dimensions of tessellation-based residue classification MEthOdS.4 Characteristics of the 1321 proteins hydrated by SOLVATE.5 Histogram of the separation distance between crystal and SOLVATE water 3010:2077.
Separation distance between DOWSER HOH placements, and the nearest crystal and SOLVATE HOH locations for 9 selected proteins.7 The effect of isolated water groups on separation between Dowser and SOLVATE water ÏOCAfÏOTIS.- - G6 cọ TH ng ng 904 37 Figure 2.8 Comparison ofresidue coordination number for full and bulk hydrations.9 Cross comparison oftwo methods for classifying residue location using Ji: hố.10 Comparison of water group residue classification methods for variations in residue location and hydratiOn. - -- - - - < cọ HH nọ 9g ng và 44 Figure 2.11 Characteristics of surface, undersurface and buried residues classified by simplex face matching. - -- cọ HT TH ng ng T00 001 0e 46 Figure 2.12 Characteristics of core and surface residues classified by water group.13 Comparison of topological parameters for human transthyretin.14 Relationship between the water group parameter and the residue water coordination NUMDBET Tati0.15 Distribution of the water group parameter for each residue type in the reference PFOteIN S€(. - Góc HH HH TH 000 688904 90 36 Figure 2.16 Comparison of mean residue values of water group parameter and of relative accessible surface area with various hydrophobicity scales.1 Distribution plots of statistical potential funetions.2 Comparison of hydrated and unhydrated potential functions.3 Comparison of the decoy discrimination scoring index for four decoy groups using four potential function strategies.4 Effect of hydrogen addition before hydration.1 Potential profiles for the wild type study proteins using the carbon-alpha Statistical pOterntiaÌ.
«cọ KT 0 119 Figure 4.2 Sensitivity of the unhydrated T4 lysozyme residual profile to single point MUTATIONS.ccccsssccccsssrcccessscccssseccesessneteeessecessssaceeessdeeesessueceessseesesessseneeessenaes 121 ix Figure 4.3 Correlation of non-conservative mutant stability change with average mutant residual for staphylococcal nuclease using three potential functions.4 Correlation of grouped mutant stability change with mean residual for T4 lysozyme for four data subsets with different pH values.5 Correlation of stability with potential for mutants of hydrophobic residues in Staphylococcal nucÌease.- - c- c1 TH TH TT c0 T404 re 132 Figure 4.6 Correlation of binned stability with average residual potential for 56 mutants of staphylococcal nuclease that make or break a hydrogen bond.7 Correlation of binned stability with average residual potential for 22 mutants of staphylococcal nuclease surface positions that reverse or neutralize a surface CHATỠC. HT TH Ti 0 cà 00 1909 081.8 Correlation of residual potential with stability change for amino acid scan at position Y27 in staphylococcal nucÏease.9 Residual potential correlated with accessible surface area for 96 mutants of ionizable residues in staphylococcal nucÌease.10 Performance comparison of three machine learning methods used to predict Stabl]ÏẨW. - HH TH TH HT no 00000 100194 104K 140 Figure 4.11 Comparison of the ROC curves for three ML models and three potential FUNCTIONS .12 Effect of attribute content on classifier performance .13 Comparison using ROC curves of the residual-only attribute set with the full attribute S€T. - Gì TH HH HH TT 011 kh 145 Figure 4.14 Effect of number of output classes on model performance .15 Measured versus predicted stability change using a support vector Tegression MOE] .16 Support vector attribute coefficients compared to wild type potential (01-12 .17 Prediction of transthyretin mutant amyloid behavior based on the RF model using CA residuals .- - cọ TH Họ 9 156 Figure 4.18 Prediction of transthyretin mutant amyloid behavior based on the RF model using SP residuaÌs.
nọ TK 000000090 157 LIST OF ABBREVIATIONS ASA accessible surface area AUC area under the ROC curve CA carbon alpha CM center of mass CMP comprehensive mutational profile CNR coordination number ratio CV circular variance DT decision tree ML machine learning PDB Protein Data Bank random forest ROC receiver operating characteristic SP split potential SV support vector WG water group wild-type ABSTRACT ADAPTATION AND USE OF FOUR-BODY STATISTICAL POTENTLAL TO EXAMINE THERMODYNAMIC PROPERTIES OF PROTEINS Gregory M. University of Cincinnati, 1969 Thesis Director: Dr. Vaisman While most proteins in biological systems are inherently stable as a prerequisite to performing their functions, a small number of normally well-behaved proteins can engage in a process of aggregation that eventually leads to the formation of an insoluble material identified as an amyloid. Details of the aggregation process are not fully known, but for some model proteins the process can be initiated with known destabilizing conditions.
While no sequence or structural similarities have been observed among the proteins, structural instability associated with a characteristic motif in the protein could be a common thread. The proposed strategy to search for such a feature employs a knowledge-based tool that examines the sequence-structure relationship in a specific target protein based on similar relationships drawn from a large representative sample of proteins. The tool incorporates a computational structural analysis known as tessellation to identify small geometric elements each containing four neighboring amino acid residues, and builds a potential score for the protein based ona statistical analysis of the appearance of these quadruplets in the reference set.