Towards a framework for multi-level modelling in Computational Biology

Size: px

Start display at page:

Download "Towards a framework for multi-level modelling in Computational Biology"

Louisa Simpson
5 years ago
Views:

1 Towards a framework for multi-level modelling in Computational Biology Sara Montagna sara.montagna@unibo.it Alma Mater Studiorum Università di Bologna PhD in Electronics, Computer Science and Telecommunications - XXIII Ciclo Second Year Ending Seminar Bologna, Italy, October 14, 2009 Montagna (UniBo) Multi-level Models in Computational Biology 14/10/ / 33

2 Outline 1 Context Computational Biology for modelling and simulating 2 Problem modelling the hierarchical organisation of biological systems 3 Contribution a multilevel modelling framework computational model specification language simulation engine 4 Evaluation analysis of Drosophila Melanogaster as a case study 5 Further application synthesis of an artificial system 6 Up-coming works Montagna (UniBo) Multi-level Models in Computational Biology 14/10/ / 33

3 Outline 1 Computational Biology 2 Modelling the Hierarchy of Biological Systems 3 A Multi-level Modelling Framework 4 Evaluation on the Drosophila Melanogaster Morphogenesis 5 Further Application Towards the Synthesis of Artificial Systems 6 Up-coming Work Montagna (UniBo) Multi-level Models in Computational Biology 14/10/ / 33

4 Computational Biology Computational biology is an interdisciplinary field that applies the techniques of computer science, applied mathematics and statistics to address biological problems. By these means it addresses scientific research topics with their theoretical and experimental questions without a laboratory. Molecular structural modelling kinetics and thermodynamics of protein functions Protein structure prediction Interpretation, classification and understanding of biological data-sets DNA, RNA, or protein sequences Modelling and simulating biological processes gene networks, metabolic pathways, tissue patterning Montagna (UniBo) Multi-level Models in Computational Biology 14/10/ / 33

5 Outline 1 Computational Biology 2 Modelling the Hierarchy of Biological Systems 3 A Multi-level Modelling Framework 4 Evaluation on the Drosophila Melanogaster Morphogenesis 5 Further Application Towards the Synthesis of Artificial Systems 6 Up-coming Work Montagna (UniBo) Multi-level Models in Computational Biology 14/10/ / 33

6 Biological Systems as Complex Systems 1 Non-linear dynamics and feedbacks interactions among components 2 Openness interaction among components and the environment 3 Hierarchy (different level of structural organization) from sequences, molecules... to pathways (such as metabolic or signaling) and networks (collection of cross-interacting pathways)... to cells, tissues, organs 4 Emergent phenomena Montagna (UniBo) Multi-level Models in Computational Biology 14/10/ / 33

7 An Example: On the Morphogenesis of Living Systems Developmental Biology researches the mechanisms of development, differentiation, and growth in animals and plants at the molecular, cellular, and genetic levels. Animal developmental steps 1 Fertilisation of one egg 2 Mitotic division 3 Cellular differentiation diverse gene expression 4 Morphogenesis control of the organised spatial distribution of the cell diversity Montagna (UniBo) Multi-level Models in Computational Biology 14/10/ / 33

mechanisms signalling pathways cell-to-cell direct interaction short and long range signals (morphogenes) interplay between cells internal activity and cell-to-cell interactions Figure by: [1]

8 Each region of the developing organism expresses a given set of genes Figure: Drosophila M. segments Figure: Zebrafish regionalisation Developmental Biology recognise as important actors in the emergence of embryonic patterning self-organised structures transcriptional control mechanisms signalling pathways cell-to-cell direct interaction short and long range signals (morphogenes) interplay between cells internal activity and cell-to-cell interactions Figure by: [1] On-line [2] An Automatic Quantification and Registration Strategy to Create a Genetic Expression Atlas of Zebrafish Embryogenesis. C. Castro et all. Accepted at IEEE Engineering in Medicine and Biology Society (EMBC 09) Montagna (UniBo) Multi-level Models in Computational Biology 14/10/ / 33

9 A Computational Model for Addressing these Scenarios Computational model requirements 1 Multi-compartment / multi-level model for reproducing the interactions and integrations of the systems components at cellular and intracellular level 2 Diffusion / Transfer for studying the effects of short and long range signals for modelling the compartment membrane 3 Stochasticity for capturing the aleatory behaviour characteristic of those systems involving few entities Montagna (UniBo) Multi-level Models in Computational Biology 14/10/ / 33

10 Brief Survey on Existing Frameworks Systems Biology Tools: See Based on classical mathematical models (ODE or PDE) CellDesigner, CellWare, COPASI, Dizzy, JDesigner, Virtual Cell,... mainly thought for intracellular aspects (biochemical pathways) do not scale well with the system complexity (e.g. number of cells) Computational Systems Biology Tools Based on computational models (stochastic process-algebras, Petri-Nets) SPIM (stochastic π-calculus), BlenX, Bio-PEPA ground on Gillespie s characterisation of chemistry as CTMC promote a view of molecules as concurrent processes recent preliminary extensions towards multi-compartimentalisation Montagna (UniBo) Multi-level Models in Computational Biology 14/10/ / 33

11 Outline 1 Computational Biology 2 Modelling the Hierarchy of Biological Systems 3 A Multi-level Modelling Framework 4 Evaluation on the Drosophila Melanogaster Morphogenesis 5 Further Application Towards the Synthesis of Artificial Systems 6 Up-coming Work Montagna (UniBo) Multi-level Models in Computational Biology 14/10/ / 33

12 A Multi-level Modelling Framework Framework s Conceptual Parts 1 Computational Model: Graph of compartments, with transfer reactions 2 Surface Language: Systems as logic-oriented description programs system structure inner chemical behaviours 3 Simulation Engine: Known O(logN) version of Gillespie SSA reproducing the exact chemical evolution/diffusion of substances Montagna (UniBo) Multi-level Models in Computational Biology 14/10/ / 33

13 Part 1: the Computational Model Structure A multi-compartment version of standard CMSB view A system is a graph-like network of compartments Each compartment hosts a chemical solution Mobility and mitotic division will be supported in future versions Montagna (UniBo) Multi-level Models in Computational Biology 14/10/ / 33

14 Part 1: The Computational Model Chemical Transfer Transfer Model Some chemical reactions can produce so-called firing molecules They are sent to a neighbouring compartment picked probabilistically Montagna (UniBo) Multi-level Models in Computational Biology 14/10/ / 33

15 Part 2: The Surface Specification Language Structure A program as a set of logic declarations Declaring molecules, reactions, compartments, links Configuring the simulation (initial state and parameters) Stating output commands Declarations can have variables and be equipped with fully expressive preconditions, acting as constraints on declarations Compiler Supporting flexibility Specifications are parsed by a Prolog interpreter An intermediate file of commands is created which will feed the simulation engine Montagna (UniBo) Multi-level Models in Computational Biology 14/10/ / 33

16 Example specification Diffusion of a substance into a grid-like tissue const size(20). molecule M where (M in [pump,field]). reaction r(pump) : [pump] --> [pump,field] rate reaction r(diff) : [field] --> [field,firing(field)] rate 0.2. reaction r(decay) : [field] --> [] rate 0.1. compartment c(x,y) where (const size(n), X in {1..N}, Y in {1..N}). link c(x,y) >>> c(x,y1) molecule field rate where ( Y1 in [Y-1,Y+1] ). link c(x,y) >>> c(x1,y) molecule field rate where ( X1 in [X-1,X+1] ). concentration 1 of pump into c(m,m) where (const size(n), M is N//2). place AnyReaction into AnyCompartment. final_steps sample_steps 100. out [molecule(c(x,y),field),delimiter] where ( const size(n), inspect(compartment c(x,y)), (Y=N -> Delimiter=end_of_line ; Delimiter=space) ). Montagna (UniBo) Multi-level Models in Computational Biology 14/10/ / 33

17 Part 3: The Simulation Engine Under the hood Main Elements An available chemical reaction in a compartment is picked up probabilistically based on its rate selected in O(logN) time, via a special binary search tree The transition modifies a small portion of data structures The transition duration is drawn with exponential distribution exactly modelling chemical dynamics according to [1] Based on: [1] D. T. Gillespie. Exact stochastic simulation of coupled chemical reactions. J. Phys. Chem., 81(25), [2] M. A. Gibson and J. Bruck. Efficient exact stochastic simulation of chemical systems with many species and many channels. J. Phys. Chem. A, 104(9), Montagna (UniBo) Multi-level Models in Computational Biology 14/10/ / 33

18 Part 3: The Simulation Engine Output (1) Produces a textual result from out commands Montagna (UniBo) Multi-level Models in Computational Biology 14/10/ / 33

19 Part 3: The Simulation Engine Output (2) Produces a textual result from out commands Montagna (UniBo) Multi-level Models in Computational Biology 14/10/ / 33

20 Part 3: The Simulation Engine Output (3) Produces a textual result from out commands Montagna (UniBo) Multi-level Models in Computational Biology 14/10/ / 33

21 Part 3: The Simulation Engine Drawing Charts Charting using any existing tool (Matlab, gnuplot,..) Montagna (UniBo) Multi-level Models in Computational Biology 14/10/ / 33

22 Outline 1 Computational Biology 2 Modelling the Hierarchy of Biological Systems 3 A Multi-level Modelling Framework 4 Evaluation on the Drosophila Melanogaster Morphogenesis 5 Further Application Towards the Synthesis of Artificial Systems 6 Up-coming Work Montagna (UniBo) Multi-level Models in Computational Biology 14/10/ / 33

23 Goals of the Model 1 To obtain a self-organised patterning such as the D.M segments regions with a different genetic expression (plig1 and plig2) boundary regions with a gradient of activity of the expressed genes Protein Number plig2 plig1 Space 2 To demonstrate the fundamental dependency between the intracellular dynamics and the extracellular ones For more details see: [1] S. Montagna and M. Viroli A computational framework for modelling multicellular biochemistry 2009 IEEE Congress on Evolutionary Computation (CEC 2009), May 2009 Montagna (UniBo) Multi-level Models in Computational Biology 14/10/ / 33

Working Hypotheses of the Model The maternal products activate either plig1 or plig2 Two regions where only one of the two pathways is active Two signalling pathways composed by 1 Receptors prec1 &

24 Working Hypotheses of the Model The maternal products activate either plig1 or plig2 Two regions where only one of the two pathways is active Two signalling pathways composed by 1 Receptors prec1 & prec2 2 Ligands plig1 & plig2 3 Intracellular cascades that transmit the signal cascade1 & cascade2 4 Genes lig1 & lig2 plig1 enhances itself production while inhibits the production of plig2 plig2 enhances itself production while inhibits the production of plig1 Montagna (UniBo) Multi-level Models in Computational Biology 14/10/ / 33

First Experiment pre-imposed regionalisation Initial configuration 4500 cells forming a 30x150 grid Few copies of the two ligands already regionalised at the beginning of zygotic expression

25 First Experiment pre-imposed regionalisation Initial configuration 4500 cells forming a 30x150 grid Few copies of the two ligands already regionalised at the beginning of zygotic expression (left) Discussion The initial pattern is strongly reinforced (right) thanks to mutual inhibition + local reinforcement Montagna (UniBo) Multi-level Models in Computational Biology 14/10/ / 33

Second Experiment spontaneous regionalisation Initial configuration 4500 cells forming a 30x150 grid same number of ligands uniformly placed Discussion regions tend to form

26 Second Experiment spontaneous regionalisation Initial configuration 4500 cells forming a 30x150 grid same number of ligands uniformly placed Discussion regions tend to form (left) and consolidate (right) dynamics is not very stable: Some further mechanism is needed Montagna (UniBo) Multi-level Models in Computational Biology 14/10/ / 33

27 Outline 1 Computational Biology 2 Modelling the Hierarchy of Biological Systems 3 A Multi-level Modelling Framework 4 Evaluation on the Drosophila Melanogaster Morphogenesis 5 Further Application Towards the Synthesis of Artificial Systems 6 Up-coming Work Montagna (UniBo) Multi-level Models in Computational Biology 14/10/ / 33

28 Pervasive Computing Scenarios Distributed infrastructures large-scale distribution opennes context-awareness self-organisation and self-adaptation How to support and engineer this scenario? Montagna (UniBo) Multi-level Models in Computational Biology 14/10/ / 33

29 A nature-inspired coordination model Mapping individuals into biochemical species One individual is modelled as a chemical substance into a solution An individual activity level is modelled as the concentration of the chemical substance an integer value measuring the strenght or health of the individual Mapping the space into the network One location is modelled as a well-mixed chemical solution of a compartment it occurs within the boundary of a biological membrane it contains several substances, modelling interacting individuals The whole system is modelled as a network-like structure of compartments individuals move to the neighbourhood as chemical substances cross membranes Montagna (UniBo) Multi-level Models in Computational Biology 14/10/ / 33

An Example: Creating a Service-Request Path Self-adaptation to topology and their changes The path adapts to the introduction of an obstacle For more details see: [1] M. Viroli and M. Casadei.

30 An Example: Creating a Service-Request Path Self-adaptation to topology and their changes The path adapts to the introduction of an obstacle For more details see: [1] M. Viroli and M. Casadei. Biochemical Tuple Spaces for Self-Organising Coordination. Coordination Languages and Models, LNCS 5521, jun 2009 [2] M. Viroli, F. Zambonelli, M. Casadei and S. Montagna. A Biochemical Metaphor for Developing Eternally Adaptive Service Ecosystems.24th Annual ACM Symposium on Applied Computing (SAC 2009), March, Montagna (UniBo) Multi-level Models in Computational Biology 14/10/ / 33

31 Outline 1 Computational Biology 2 Modelling the Hierarchy of Biological Systems 3 A Multi-level Modelling Framework 4 Evaluation on the Drosophila Melanogaster Morphogenesis 5 Further Application Towards the Synthesis of Artificial Systems 6 Up-coming Work Montagna (UniBo) Multi-level Models in Computational Biology 14/10/ / 33

32 Up-coming Work Agenda Tool side re-engineer the tool towards a community release support dynamic networks and mitotic division improve chemical transfer model perform parameter tuning Biological systems model side analyse Zebrafish morphogenesis using real biological data (not still fully available) inferring the gene networks from literature Artificial design side program clustering, gradients, paths, self-composition find applications for patterns of natural/artificial biochemistry implement real-life systems: Pervasive Displays Infrastructure Montagna (UniBo) Multi-level Models in Computational Biology 14/10/ / 33

33 Towards a framework for multi-level modelling in Computational Biology Sara Montagna sara.montagna@unibo.it Alma Mater Studiorum Università di Bologna PhD in Electronics, Computer Science and Telecommunications - XXIII Ciclo Second Year Ending Seminar Bologna, Italy, October 14, 2009 Montagna (UniBo) Multi-level Models in Computational Biology 14/10/ / 33

SPA for quantitative analysis: Lecture 6 Modelling Biological Processes

1/ 223 SPA for quantitative analysis: Lecture 6 Modelling Biological Processes Jane Hillston LFCS, School of Informatics The University of Edinburgh Scotland 7th March 2013 Outline 2/ 223 1 Introduction