Multi-Layer Modeling in Systems Biology Sriram Iyengar, PhD, Associate Professor, School of Biomedical Informatics Associate Professor, School of Biomedical.

Multi-Layer Modeling in Multi-Layer Modeling in Systems BiologySystems Biology

Sriram Iyengar, PhD,Sriram Iyengar, PhD, Associate Professor, School of Biomedical InformaticsAssociate Professor, School of Biomedical InformaticsUniversity of Texas, HoustonUniversity of Texas, Houston

Biological processes involved in diseases and disease progression require integration of multiple types of knowledge.

2

A goal of Systems BiologyA goal of Systems Biology

t = time

Functionality(t) = (y1,y2,y3,..,t) = f(x1,x2,x3,…, t)

Other goals, maybe on the way to achieving the above: Enhance detailed understanding of biological processes

3

Biology is complexBiology is complex!

Interactome - Protein-protein Interaction Map- The Scientist 2004, 18(12):18

4

What can systems biology do for What can systems biology do for medicine and patient caremedicine and patient care

Develop accurate specific and personalized models of actual biological processes based on tissue samples and individual genetics

Generate hypotheses for disease presence and disease progression

Support personalized medicine Develop targeted therapies. Support translational medicine

5

The biological cell: Nature’s canonical The biological cell: Nature’s canonical object in the sense of Object oriented object in the sense of Object oriented programming programming “The fundamental unit of life is the biological cell and hence

should be the focus for current research in digital biology”- Sidney Brenner, 2002 Nobel laureate in medicine

Cellular process modeling is now an active research area- E-cell project (Japan)- Virtual cell project (National Resource for Cell Analysis and

Modeling, USA)

Create in-silico digital cell

6

But: Biological cells are extremely But: Biological cells are extremely intricateintricate ~500 types of cells

- red blood cells, neurons, and more

>60 billion cells in the human body- >100 billion Symbiotic bacteria

30,000 proteins expressed at any given time by a cell That’s only a fraction of what a cell does, or what it looks

like A cubic cm of the human brain can contain 50 million

neurons, each supported by 10 glial cells and connected to many other neurons

7

In silico cell modelingIn silico cell modeling

The challenge- Creating an executable software description of a

biological cell is an extremely formidable task

The computer science response- Reduce complexity

Use artificial Intelligence and software engineering methods

8

Why do this? MorphoproteomicsWhy do this? Morphoproteomics

Development of malignant tumors depends on- Timing

Cell cycle phase- Location

Interactions in cell- Pathways

Signal cell proliferation- Crosstalk

Messages sent across multiple signaling pathways Robert Brown, MD, Deputy Chief of Pathology, Univ. Texas,

Houston

9

Morphoproteomics visual analysisMorphoproteomics visual analysisof stained tumor cellsof stained tumor cells

10

Glioblastoma multiformeGlioblastoma multiforme

11

Implications for cancer therapyImplications for cancer therapy

Multidimensional approach required to analyze- Timing- Location- Signals- Crosstalk

Cell morphology - Where are particular signal transduction events taking

place?- How does intracellular morphology, eg cytoskeleton,

impact signaling, cross-talk and ultimately, disease progression?

12

Techniques to reduce complexityTechniques to reduce complexity

Abstraction- Capture the essential or defining attributes of a body of

knowledge

Modularity- Divide and conquer

Reduce complex problems into simpler sub-problems “Solve” each sub-problem by domain specialists

13

Model Abstraction & AssemblyModel Abstraction & Assembly Abstraction

- Simplify the system in such a way that the properties of interest are preserved

- Einstein: Make the abstracted system “as simple as possible, but no simpler”

Assemble the sub-problems- “Glue” via a suite of interfaces (APIs - Application

Programming Interfaces)

Create a multi-layer model: what criteria? Physical scale? Functional behavior?

Injection of humility: remember that ‘All models are wrong, but some are useful’… GEP Box (a statistician)

14

Example Multi-layer modelExample Multi-layer model

ISO-OSI - International Standards Organization - Open Systems

Interconnect

- Used to model and create systems for data communications

- Email, web, file transfer

15

ISO-OSI model for Data ISO-OSI model for Data CommunicationsCommunications

16

Multi-layer Modeling benefitsMulti-layer Modeling benefits

Complex problem reduced to tractable sub-problems- Data homogeneous within a layer

Each layer can be dealt with by appropriate specialists- Problems in the physical layer differ from those in the

application layer- Different skills needed- No one person must be expert in several disciplines (layers)

Universal focus in thought and discussion- “TCP belongs to transport layer”

Everyone knows what TCP does and does not do

17

Building multi-layer models to Building multi-layer models to support Systems Biologysupport Systems Biology How can this be done?

Let’s consider one aspect at a time- Simplify the system via divide & conquer

Define the sub-problems and their interfaces- Data- Processes- Interactions

Goal - To create a model based on the synthesis of heterogeneous

knowledge Biological cell seems to be a good place to start

18

Multi-layer Multi-layer Cell ModelCell Model

____Metabolism /____Metabolism / SignalingSignaling

____Physical Structure____Physical Structure

____Symbolic Biochemistry____Symbolic Biochemistry

____Biochemistry____Biochemistry

____Functionality____Functionality

____Molecular Morphology____Molecular Morphology

19

Molecular MorphologyMolecular Morphology

Bio-molecular structure & form Molecular modeling Molecular dynamics

- UIUC group, www.ks.uiuc.edu Protein folding, conformation Interactions at molecular level

- UTH -www.biomachina.org

20

Biochemistry (Thermodynamics)Biochemistry (Thermodynamics)

View cell as a dynamic chemical machine Use reaction kinetics, rate constants

Michaelis-Menten equation Set multiple equilibria Identify bio-chemical minutiae by which organic molecules

interact with each other Examples

Plateki (www.biokin.com) Determines inhibition constants from plate-reader

data

21

Symbolic biochemistrySymbolic biochemistry

Abstract symbolic models of biochemical assemblies from components- DNA as an assembly of A, G, C, T- Proteins as assemblies of amino

acids Genetic code

- Human genome project Algorithms:

Smith-Waterman, Hidden Markov models

22

MetabolismMetabolism

Cells convert nutrients into energy via metabolic pathways & processes

Symbolic models facilitate understanding of cell’s “Quality of life”

Is cell functioning at 100% energy efficiency? Examples

- Biocyc projects- E-cell project

23

SignalingSignaling

Processes by which cell communicates with external entities and intra-cellular components

Signal transduction pathways

Examples- Pathway Logic- Pi Calculus- Rewriting Logic and Protein

Functional Domains

24

Physical Structure Cell MorphologyPhysical Structure Cell Morphology

Cell Assembly- anatomic components

3-D shapes 3-D spatial relationships

Physical characteristics of cell components- Texture- Permeability

Specialized components- Axons, Dendrons etc

Cytoview Project: In collaboration with Indian Inst of Science Journal of Biosciences, v32 #5 “Cytoview: Development of a cell modeling

framework.

25

Cytoview: Ontology of componentsCytoview: Ontology of components

26

Relationship MatrixRelationship Matrix

27

FunctionalityFunctionality

What does cell do?- Use mathematical and software

idealizations- Black box, systems approach- Like ISO-OSI application layer

Examples- McCulloch-Pitts model of neuron- Models of optical cells

The challenge of in silico cell modeling- How to relate functionality to “rest of cell”

28

InterfacesInterfaces

Receptors & other chemical entities- Enable signaling- Serve as interfaces among pathway segments

Receptors- From computing / communications view

Receptors are “control hubs” They determine topologies of signaling and

metabolic pathways- Receptor chemistry is an important discipline in itself

Symbolic descriptions of signaling interfaces help define cell processes

29

Multi-layered synthesis of Multi-layered synthesis of heterogeneous knowledgeheterogeneous knowledge

It is most likely impossible that - The complete story

From functionality down to the molecular layer - Can be expressed by a set of mathematical equations

But it is possible that - Algorithmic procedures encoded in software can do so

The layered model is intended to provide a framework for developing software & underlying knowledge bases- Morphoproteomics- Other applications ???

30

ExampleExample Firing of neuron can be

represented as- A cascade of events & processes across layers

Depending on desired level of detail, a multi-layer software model of neuron firing may -instantiate objects, databases, & software from several levels

Functionality is clearly dependent on physical structure -Axons, dendrites

31

ExampleExample

Neuron’s continued existence & wellness depends on metabolism

Control behavior is mediated by signal transduction-Biochemical neural signals at input synapses

Biochemistry & molecular dynamics mediate actual passage of electrical signals (Hodgkin-Huxley)

32

Layered model tasksLayered model tasks Define overall

- Application Programming Interfaces, or rather meta- APIs, which describe how layers should communicate with each other

Define for each layer- Uniform standards for representations of knowledge- Standards for databases and database access

Note: above tasks are simplified since there is greater homogeneity of knowledge within a layer

Interactions between layers: eg, can results derived from reasoning in the symbolic biochemistry layer be used to reduce dimensionality or enhance computational tractability in the biochemistry layer

33

SummarySummary

Bio-molecular complexity may be reduced by- Abstraction- Modularization

The goal of creating an in silico complex biological objects, eg, biological cell, may be facilitated by means of the layered model described

Multi-layer model as an organizing principle to guide discussion, research, development?

34

Challenges Challenges

To date most methods have been used for analysis of interactions within a single cell

In reality, cells work together - or aqainst each other - in biological processes

Cancer involves millions of cells and inter-cellular interactions (metastasis)

Computer science can offer tools to solve these challenging problems

Morphoproteomics, with its depth of heterogeneous knowledge, is a unique application area that bridges “bench to bedside” : Translational, personalized medicine

35

AcknowledgementsAcknowledgements

Mary McGuire, MS, PhD, UT Houston Jack W Smith, MD, PhD, UT Houston Robert Brown, MD, UT Houston Pat Lincoln, PhD, SRI Carolyn Talcott, PhD, SRI David Mercer, MD, University of Nebraska Suma Chandra, PhD, Indian Institute of Science N. Balakrishnan ,PhD, Indian institute of Science