Estimation of Transformed Reaction Gibbs Energy for Thermodynamically Constraining Metabolic Reaction Networks
Total Page:16
File Type:pdf, Size:1020Kb
Estimation of Transformed Reaction Gibbs Energy for Thermodynamically Constraining Metabolic Reaction Networks Hulda S. Haraldsdóttir FacultyFaculty of of Industrial Industrial Engineering, Engineering, MechanicalMechanical Engineering Engineering and and Computer Computer Science Science UniversityUniversity of of Iceland Iceland 20142014 ESTIMATION OF TRANSFORMED REACTION GIBBS ENERGY FOR THERMODYNAMICALLY CONSTRAINING METABOLIC REACTION NETWORKS Hulda S. Haraldsdóttir Dissertation submitted in partial fulllment of Philosophiae Doctor degree in Bioengineering Advisor Dr. Ronan M. T. Fleming Thesis Committee Prof. Sigurður Brynjólfsson Dr. Ines Thiele Prof. Jón J. Jónsson Opponents Prof. David A. Fell Dr. Ross P. Carlson Industrial Engineering, Mechanical Engineering and Computer Science School of Engineering and Natural Sciences University of Iceland Reykjavik, January 2014 Estimation of Transformed Reaction Gibbs Energy for Thermodynamically Con- straining Metabolic Reaction Networks 0 Estimation of DrG in Metabolic Reaction Networks Dissertation submitted in a partial fulfillment of a Ph.D. degree in Bioengineering Copyright © 2014 Hulda S. Haraldsdóttir All rights reserved Industrial Engineering, Mechanical Engineering and Computer Science School of Engineering and Natural Sciences University of Iceland Hjarðarhagi 2-6 107, Reykjavik Iceland Telephone: 525 4700 Bibliographic information: Hulda S. Haraldsdóttir, 2014, Estimation of Transformed Reaction Gibbs Energy for Thermodynamically Constraining Metabolic Reaction Networks, Ph.D. thesis, Faculty of Industrial Engineering, Mechanical Engineering and Computer Science, University of Iceland. ISBN 978-9935-9146-4-4 Printing: Háskólaprent, Fálkagata 2, 107 Reykjavik Reykjavik, Iceland, January 2014 Abstract Constraint-based modeling enables investigation of the properties and functions of metabolic reaction networks through prediction of steady-state network fluxes that are feasible under a set of physicochemical constraints. Application of thermody- namic constraints has been hindered in part by lack of data on the key thermody- namic parameter; transformed reaction Gibbs energy. This thesis focuses on devel- opment of computational methods for estimating transformed reaction Gibbs ener- gies in metabolic network reconstructions. Methods for estimating transformed reaction Gibbs energies existed at the outset of this thesis. The limitations of these methods were revealed through their application to the genome-scale human metabolic network reconstruction Recon 1. Subsequent efforts aimed to overcome some of these limitations through development of im- proved estimation methods. Estimation of transformed reaction Gibbs energy can roughly be broken down into two steps: 1) estimation of standard reaction Gibbs energy, and 2) adjustment to in vivo conditions. A novel estimation framework, called the component contribution method, was developed as part of this thesis to tackle the former step. An existing application for performing the latter step, called von Bertalanffy, was continually updated throughout this thesis to reflect findings of what factors contribute to the accuracy of estimated transformed reaction Gibbs energies. The latest version of von Bertalanffy is fully integrated with an implementation of the component con- tribution method. The most time-consuming step in applying the estimation meth- ods developed in this thesis is collection of necessary input data, such as metabolite structures. Strategies for partially automating collection of structures are proposed. Útdráttur Skorðuð líkön má nota til að rannsaka eiginleika og virkni efnaskiptaneta, með því að spá fyrir um stöðug ástönd sem slík net geta náð innan gefinna eðlisefnafræðile- gra skorða. Varmafræðilegum skorðum hefur aðeins verið beitt á efnaskiptalíkön að takmörkuðu leiti, meðal annars vegna skorts á gögnum um ummyndaða Gibbs orku efnaskiptahvarfa. Þessi ritgerð fjallar um þróun reiknifræðilegra aðferða til að meta ummyndaða Gibbs orku fyrir hvörf í líkönum af efnaskiptanetum. Aðferðir til að meta ummyndaða Gibbs hvarforku voru til þegar vinna við þessa ritgerð hófst. Þegar þessum aðferðum var beitt á líkan af efnaskiptaneti manna, sem kallast Recon 1, komu ýmsar takmarkanir þeirra í ljós. Bættar matsaðferðir voru þróaðar í kjölfarið til að yfirstíga sumar af þessum takmörkunum. Mati á ummyndaðri Gibbs hvarforku má skipta gróflega niður í tvö skref: 1) mat á staðlaðri Gibbs hvarforku, og 2) aðlögun að skilyrðum í lifandi verum. Til að takast á við fyrra skrefið var þróuð nýstárleg matsaðferð sem kallast þáttunaraðfer- ðin. Fyrir seinna skrefið var forritið von Bertalanffy uppfært í takt við niðurstöður um hvaða þættir hefðu mest áhrif á nákvæmni mats á ummyndaðri Gibbs orku. Nýjasta útgáfa von Bertalanffy er fyllilega samþætt útgáfu af þáttunaraðferðinni. Tímafrekasti þátturinn í að beita matsaðferðunum sem hér voru þróaðar er að safna nauðsynlögum gögnum, svo sem gögnum um byggingu sameinda. Stungið er upp á aðferðum til að sjálfvirknivæða söfnun gagna um sameindabyggingu að einhverju leyti. Dedication To Valur. Þú ert langbestur. Contents List of Figures xi List of Tables xiii List of Symbols xv Acknowledgements xvii 1. Introduction 1 1.1. Metabolic network reconstructions . .3 1.2. Constraint-based modeling . .5 1.2.1. Flux balance analysis . .6 1.3. Biochemical reaction Gibbs energy . .7 1.4. Thermodynamic constraints . 11 1.4.1. Reaction directionality . 11 1.4.2. Energy balance . 12 1.4.3. Incorporation of metabolite concentrations . 14 1.5. Estimation of reaction Gibbs energy . 16 1.5.1. Estimation of standard reaction Gibbs energies . 17 1.5.2. Estimation of transformed reaction Gibbs energies . 20 1.6. Summary of main contributions . 20 1.6.1. Paper 1 . 20 1.6.2. Paper 2 . 23 1.6.3. Paper 3 . 24 2. Quantitative assignment of reaction directionality in a multicom- partmental human metabolic reconstruction 27 2.1. Abstract . 27 2.2. Introduction . 27 2.3. Methods . 29 2.3.1. Standard transformed metabolite Gibbs energy of formation 29 2.3.2. Standard transformed reaction Gibbs energy . 32 2.3.3. Metabolite concentrations . 34 2.3.4. Quantitative assignment of reaction directionality . 34 2.3.5. Concentration variability . 35 2.4. Results and discussion . 36 2.4.1. Standard transformed Gibbs energy of formation . 36 vii 2.4.2. Standard transformed reaction Gibbs energy . 37 2.4.3. Quantitative assignment of reaction directionality . 38 2.4.4. Concentration variability . 41 2.5. Conclusions . 43 3. Consistent estimation of Gibbs energy using component contribu- tions 47 3.1. Abstract . 47 3.2. Introduction . 47 3.2.1. Unifying reactant and group contribution methods . 50 3.3. Results . 53 3.3.1. The component contribution method . 53 3.3.2. Validation results . 59 3.3.3. Application to genome-scale metabolic reconstructions . 60 3.4. Discussion . 63 3.5. Methods . 65 3.5.1. Calculation of confidence intervals . 65 3.5.2. Leave-one-out cross validation . 67 3.5.3. Calculation of prediction intervals . 67 3.5.4. Adjustment to in vivo conditions . 68 3.5.5. Implementation and availability of code . 68 4. Comparative evaluation of open source software for mapping be- tween metabolite identiers in metabolic network reconstructions: application to Recon 2 71 4.1. Abstract . 71 4.2. Introduction . 72 4.2.1. Applications . 75 4.3. Results . 77 4.3.1. Identifier mapping tests . 77 4.3.2. Optimization of performance . 79 4.3.3. Update of Recon 2 metabolite annotations . 81 4.4. Discussion . 83 4.5. Conclusions . 87 4.6. Methods . 88 4.6.1. Design of identifier mapping tests . 88 4.6.2. Scoring . 88 5. Future perspectives 91 5.1. Further development of methods for standard reaction Gibbs energy estimation . 91 5.2. Further development of methods for adjustment to in vivo conditions 94 5.3. Applications . 96 viii Bibliography 97 A. Supplementary material for Chapter 2 117 A.1. Cell compartment pH . 117 A.1.1. Cytosol and nucleus . 118 A.1.2. The endoplasmic reticulum, Golgi apparatus and lysosomes 118 A.1.3. Extracellular . 118 A.1.4. Mitochondria . 119 A.1.5. Peroxisomes . 119 A.2. Cell compartment electrical potential . 120 A.2.1. Mitochondria . 121 A.2.2. Extracellular fluid . 121 A.2.3. Golgi apparatus . 121 A.2.4. Endoplasmic reticulum . 122 A.2.5. Lysosomes . 122 A.2.6. Peroxisomes . 122 A.3. Uncertainty in thermodynamic data . 123 A.3.1. Uncertainty in standard transformed metabolite Gibbs en- ergy of formation . 123 A.3.2. Uncertainty in standard transformed reaction Gibbs energy . 125 A.4. Thermodynamic treatment of hydrogen ions and charge in multi- compartmental metabolic reactions . 127 A.5. Concentrations of water and dissolved gases . 132 A.6. Reactions with inconsistent qualitative and quantitative directional- ity assignments . 133 A.6.1. Reactions with incorrect quantitative directionality assign- ments . 133 A.6.2. Reactions with correct quantitative directionality assignments134 A.6.3. Reactions with potentially incorrect stoichiometries in Re- con1 ............................. 137 A.7. Supplementary tables for Section 2 . 140 A.8. Supplementary figures for Section 2 . 143 B. Supplementary material for Chapter 3 153 B.1. Training data . 153 B.1.1. Errors associated with the inverse Legendre transform . 154 B.1.2. Weighing the training observations . 154 B.2. Group decomposition . 156 B.3. Full mathematical derivation of the component contribution method 156 B.4. Estimation of error in group model . 157 P · P = P B.4.1. Proof that N (S) N (G >S) N (S) ............. 159 B.4.2. Error in current