Systems Biology. Nicolás Loira Center for Mathematical Modeling November 2014

Transcription

1 Systems Biology Nicolás Loira Center for Mathematical Modeling November 2014

2 Overview Intro to Systems Biology Metabolic Models

3 Systems Biology

4 Parts and pieces 4

5

6 Systems

7 Sub-systems

8 Systemic view of a Cell Signaling networks Regulatory networks Metabolic networks 8

9 Metabolic Networks

10 Signaling networks

11 Regulatory networks [Calzone, 2008]

12 Network Inference Network inference regression procedure a 0,i a 1,i a 2,i a 3,i a 4,i x x x x _ x x 4 + x 2 x 3

13 Scientific & Biotech Applications 1. Contextualization of data 2. Guidance of metabolic engineering 3. Directing hypothesis-driven discovery 4. Discovery of multi-species relationships 5. Network property discovery

14 Bioinfo. sub-fields -- Sequence analysis -- Genome annotation -- Computational evolutionary biology -- Literature analysis -- Analysis of gene expression -- Analysis of regulation -- Analysis of protein expression -- Analysis of molecules (metabolomics) -- Comparative genomics -- Modeling biological systems -- High-throughput image analysis -- Structural Bioinformatics

15 Tools from Math, CS, IT ---- algorithms, computational biology ---- databases & information system ---- web, web services ---- software engineering ---- HPC (High Performance Computing) ---- data mining ---- image processing ---- modeling and simulation ---- discrete mathematics (eg: graphs, logic) ---- control and system theory ---- statistics

16 engineering vs. reverse engineering biology: reverse engineering systems knowledge Reverse engineering is the process of discovering the technological principles of a device, object, or system through analysis of its structure, function, and operation. Engineering is the discipline, art, skill and profession of acquiring and applying scientific, mathematical,economic, social, and practical knowledge, in order to design and build structures, machines, devices, systems, materials and processes

17 Biology and models -- biology: reverse engineering systems models -- text, symbols, standards, language, math -- models: static vs dynamic ---- static: data and their relationships ---- dynamic: fluxes and multi-agent -- genomics -- prokaryote vs eukaryote

18 Models Descriptive Predictive Generic *choose two

19 Descriptive Models vs Black Box

20 Predictive Models experimental evidence biological model Simulate Simulation results modified model Simulate Simulation results Do experiment

21 Generic Models Specific model Generic model

22 Iterative model improvement experimental evidence biological model Simulate Simulation results Validate model accuracy

23 Metabolic Models

24 Definition of Metabolic Pathways l A chemical reaction interconverts chemical compounds (analogous to a production rule) A + B = C + D l An enzyme is a protein that accelerates chemical reactions. Each enzyme is encoded by one or more genes. l A pathway is a linked set of reactions (analogous to a chain of rules) A C E

25 Pathways

26 What is a Metabolic Pathway? A pathway is a conceptual unit of the metabolism An ordered set of interconnected, directed biochemical reactions A pathway forms a coherent unit: Boundaries defined at high-connectivity substrates Regulated as a single unit Evolutionarily conserved across organisms as a single unit Performs a single cellular function Historically grouped together as a unit All reactions in a single organism

27 Genome-scale Metabolic Networks Compartments Transports Influx Outflux Biomass Chains of reactions 27

28 Elements of Met.Networks a Chemical Reactions b Genes->Proteins->Reactions c cytosol extracellular nucl. mitocondria

29 Stoichiometry is the measuring of metabolites in a chemical reaction Reaction2: 2 M1 + 3 M2 => 1 M3 + 4 M4 ( )

30 A network of reactions can be described with an Stoichiometric Matrix R1 R2 R3 R4 R5 S = M1 M2 M3 M Systems Biology: Properties of Reconstructed Networks (Palsson, 2006)

31 Instantiation of a Reaction reaction Organism1 (Source) R1 S 1, S 2, S 3 genes Organism2 (Target) R1 T 1??? 31

32 Gene association S 1 S 2 S 3 Γ= {S 1,S 2,S 3 } β= and S 1 or Reaction R Gene association: S 1 and (S 2 or S 3 ) S 2 S 3 32

33 KEGG

34 KEGG Color Mapper

35 Current metabolic models of S.cerevisiae Model ID Publication Genes Reactions Metabolites Compart -ments iff708 (Förster, 03) 708 1, ind750 (Duarte, 04) 750 1, ill672 (Kuepfer, 05) 672 1, iin800 (Nookaew, 08) 800 1,446 1,118 4 ijm832 (Herrgård, 08) 832 1,857 2, ijm832 no compartments (Herrgård, 08) 832 1,573 1,748 2 imm904 (Mo, 09) 904 1,577 1,392 8

36 SBML beginning of model definition list of function definitions (optional) list of unit definitions (optional) list of compartment types (optional) list of species types (optional) list of compartments (optional) list of species (optional) list of parameters (optional) list of initial assignments (optional) list of rules (optional) list of constraints (optional) list of reactions (optional) list of events (optional) end of model definition

37 Visualization tools Cytoscape CellDesigner

38 Current reconstructions Genome-scale models are hard and expensive to build Most reconstructions are for bacteria There are no tools to correctly reconstruct models for eukaryotes [Oberhardt, 2009]

39 [Thiele, Nature de novo reconstruction Protocols, 2010]

40 Auto: Pathway Tools Software: PathoLogic l Computational creation of new Pathway/Genome Databases l Transforms genome into Pathway Tools schema and layers inferred information above the genome l Predicts operons l Predicts metabolic network l Predicts pathway hole fillers l Infers transport reactions [Slides by Peter Karp]

41 Pathway Tools Software: Pathway/Genome Editors l Interactively update PGDBs with graphical editors l Support geographically distributed teams of curators with object database system l Gene editor l Protein editor l Reaction editor l Compound editor l Pathway editor l Operon editor l Publication editor

42 Pathway Tools Software: Pathway/Genome Navigator l Querying, visualization of pathways, chromosomes, operons l Analysis operations l Pathway visualization of gene-expression data l Global comparisons of metabolic networks l Comparative genomics l WWW publishing of PGDBs l Desktop operation

43

44 Auto: Pantograph Re-using existing models Saccharomyces cerevisiae Saccharomyces paradoxus Saccharomyces mikatae Saccharomyces kudriavzevii Saccharomyces bayanus Saccharomyces exiguus Saccharomyces servazzii Saccharomyces castellii Candida glabrata Zygosaccharomyces rouxii Kluyveromyces thermotolerans We need methods that can use an existing model to provide a base to build models for other organisms This should decrease the amount of work needed to build a metabolic model Kluyveromyces waltii Saccharomyces kluyveri Kluyveromyces lactis Kluyveromyces marxianus Ashbya gossypii 44

45 Model Organism 1 Auto: Pantograph Genes Organism 1 ] PANTOGRAPH METHOD Model Organism 2 Genes organism 2

46 Pantograph Workflow Target annotated genes Source annotated genes Reaction Instantiator Instantiated Reactions curator editediteditedit Source metabolic model Model Instantiator target draft model Curation target model accuracy Validation target curated model experimental evidence 46

47 Reconstructions with Pantograph [Loira et al., 2012] Yarrowia lipolytica (INRA Micalis, France) Biofuels Nannochloropsis salina (AUSTRAL Biotech, Chile) Ectocarpus siliculosus (INRIA Dylis, France) Biomining Acidithiobacillus ferrooxidans (MATHomics, CMM, U.Chile)

48 Gap Filling Automatic procedures can generate models with gaps Gaps can exist because: A U U The organism doesn t have that reaction C? D Metabolic network U Candidates to fill gap The tools could not identify an homolog for the genes The organism have a different way to produce this enzymatic reaction

49 Manual Curation Models reconstructed by automatic methods still require manual curation To Fix non-obvious gaps (but obvious for an expert in the organism) To Add relevant knowledge from the literature

50 Manual curation Target annotated genes Source annotated genes Reaction Instantiator Instantiated Reactions curator editediteditedit Source model Model Instantiator target draft model Curation target model accuracy Validation target curated model experimental evidence 50

51 Simulation and predictions of metabolism Models can be used to predict behavior Model Simulate Predict Experimental Validate Data

52 Fluxes Fluxes account for the number of times a reaction Representation of the variation of Metabolites in a time unit, given some flux: S * v = dx/dt S: Stoichiometric Matrix v: Flux Vector happens x: Metabolite vector x=(m1 M2 M3 M4... )

53 Steady State Steady-state mean number of internal metabolites is constant in time S * v = dx/dt = 0 Biological Systems tend to stabilize with time in a Steady State

54 Flux Balance Analysis (FBA) Flux Balance Analysis (FBA) is useful to search fluxes that maximize a function Input Substrates Metabolic Model Look for optimal path to maximize production Cell Products Max f() (biomass, ethanol, etc.)

55 FBA solutions are restricted by constrains Constraints Model Biomass Function

56 FBA can be used to maximize biomass production Biomass = ,3betaDglcn ala amp arg asn asp atp cmp cys damp dcmp dgmp dtmp ergst gln glu gly glycogen gmp h2o his ile leu lys mannan met pa pc pe phe pro ps ptd1ino ser so thr tre triglyc trp tyr ump val zymst

57 From a constrained flux problem we build an LP problem Maximize: ω define metabolites in biomass function Subject to: α and β are bounds to fluxes

58 We use FBA to measure effects of reaction deletion in biomass Non-lethal From LP B Growth Deletion Max Biomass{< B No Growth Lethal Deletion

59 Genetic conditions (Gene KOs) Knocking out a gene and measuring growth growth (OD) Viable [Joice, 2006] (threshold=⅓ of average OD) Lethal time

60 Simulation vs experimental evidence A simulation is accurate if predicts (lack of) growth correctly Experimental Growth Growth True Positive Simulation FBA No Growth False Positive No Growth Growth No Growth False Negative True Negative

61 Metabolic model accuracy Example: Model ind750, Biomass Function ind750 Against experimental results from (Winzeler,1999) Rich Media Minimal Media Accuracy= (TP+TN) (TP+TN+FP+FN) TP TN 1 6 FP 3 4 FN 9 4 Accuracy Accuracy= (Tp+Tn)/(tp+tn+fp+fn) sensitivity= tp/(tp+fn) specificity= tn/(tn+fp) geom.mean= sqr(sens*specif)

62 Accuracy report Model Simulator MATLAB COBRA Toolbox Literature Prediction Experiment Prediction Prediction Prediction VALIDATOR Experiment Experiment Experiment Prediction Experiment Accuracy report

63 Detailed validation Y. lipolytica S. cerevisiae Gene Exp. Simul. Ref Media [36] Ethanol knocked YALI0C24101g locus ortholog YGL062W name PYC1 Growth + Growth Result - FN [36] Aspartate YALI0C24101g YGL062W PYC1 + + TP [36] Glutamate YALI0C24101g YGL062W PYC1 + - FN [36] YNBD YALI0C16885g YER065C ICL1 + + TP [36] Ethanol YALI0C16885g YER065C ICL1 - - TN [36] Aspartate YALI0C16885g YER065C ICL1 + + TP [36] Glutamate YALI0C16885g YER065C ICL1 + - FN [36] YNBD YALI0C24101g YGL062W ICL1 YALI0C16885g YER065C PYC1 - - TN [36] Ethanol YALI0C24101g YGL062W ICL1 YALI0C16885g YER065C PYC1 - - TN [36] Aspartate YALI0C24101g YGL062W ICL1 YALI0C16885g YER065C PYC1 + + TP YALI0C24101g YGL062W ICL1 63

64 Y. lipolytica Model validation 152 experiments 98 simulations 39 True Positives 16 False Positives 25 True Negatives18 False Negatives 64

65 Tools used for this analysis COBRA toolbox: for FluxBalanceAnalysis Matlab glpk/glpkmex: for solving LP libsbml/sbmltools: for SBML handling Pantograph

66 Validation and iterative improvement Target annotated genes Source annotated genes Reaction Instantiator Instantiated Reactions curator editediteditedit Source model Model Instantiator target draft model Curation target model accuracy Validation target curated model experimental evidence 66

67 FIN Nicolás Loira Center for Mathematical Modeling November 2014