© 2006 UCSF Biosystems Group 1 EO Modeling Workshop Tucson May 16, 2006 C. Anthony Hunt UCSF Biosystems Group Designing and Building In Silico Analogues

1© 2006 UCSF Biosystems Group

EO Modeling Workshop Tucson

May 16, 2006

C. Anthony Hunt

UCSF Biosystems Group

Designing and Building In Silico Analogues

of In Vitro Models


Two examples of how we would like models to be useful …

and what we have learned while making important progress

toward our goals


1. An in silico analogue of epithelial cell morphogenesis in vitro

normal & abnormal



2. An in silico analog of a liver – normal or abnormal –

that will respond to compounds

– alone or in combination –– studied & not yet studied –

analogous to real livers



Model Use In Biomedical Research


Need To Bridge The Gap


The Current State of Modeling – my view –


Current Successes

Dramatic increases in data

Inductive method hasproven verysuccessful in facilitatinghypotheses generation

Hypotheses about how the data might have been generated


Current ProblemsAnticipatedincrease inscientific

usefulnessof inductive models

has failed to materialize

Their usefulness in anticipating the consequences ofinterventions has also not improved

Why? Induction abstracts away detail: the heuristic value


Current Problems

Why have a model if you cannot predict?

Because we want to understand how (& why) the system is exhibiting the observed behavior

Prediction requires deep knowledge and rich theoriesWith a purely physical systems that is possibleYou are doing it

A liver cell is alive

There is no usable theory of life


Current Problems

We need new modeling methods …

in addition to those we have …that can help us acquire the deep knowledge & usable theories

Absent prediction, we stillneed to anticipate futurebehaviors to a useful degree

For that, we need to enable these new models withsufficient heuristic value to anticipate the consequences of interventions


Richard Feynman

• Richard Feynman (1988): “What I cannot create I do not understand”

• To understand how interventions produce system level changes in phenotype we need to create & build

System-Wide Devices

• that can exhibit some of the behaviors of interest

The best ways to learn about a phenomenon … build a device that exhibits that behavior


Richard Feynman

For a biological system I can not use the same approach

With complete knowledge of a physical system I can describe component behaviors – at what ever level of detail is needed – and thereby build an in

silico scale model


• each with their own agenda – axioms …

• & plug them together in biologically feasible ways…

• I can then measure the resulting system behaviors (as I would in a wet-lab) ...

• and compare those measures to measures of the behavior of interest

However, I can use OO programming to build independent analogues of biological components …


We call this the synthetic method of M&S to distinguish it from the inductive method

I can then iteratively adjust the in silico analogue to get improved overlap of measured

observables


Toward a Foundationfor In Silico Experimental Biology


A Foundationfor In Silico Experimental Biology


First Hard Lesson Learned

• Scientifically useful models need to be use–focused…

• not data–focused


How will the model – the analogue –

be used?

To obtain, what properties are needed?


• 2nd hard lesson: not all properties can be anticipated in advance:

• design for flexibly


Our list of properties: context: in silico liver (ISL)

EC Morph properties are essentiallyidentical


Properties Required

• Models accurately represent intrahepatic events

• Clear physiological mapping between referent and model components …

• requires model & referent observables to be consistent

• When dosed with a simulated compound, the ISL generates data that are, to a domain expert (in a type of Turing test), experimentally indistinguishable from referent wet-lab data; …

• model & framework must be suitable for experimentation


Properties Required

• These properties:to obtain, must use discrete interactions

• The ISL must be transparent: simulation details, as it progresses, need to be visualizable

• Components articulate: it must be easy to join, disconnect, and replace components

• Components can be easily reconfigured to represent different histological, physiological, or experimental conditions


Properties Required

• It must be relatively simple to change usage and assumptions, or increase or decrease detail in order to meet the particular needs of an experiment, …

• without requiring significant re-engineering of the model.

• The ISL must be reusable for simulating the clearance and metabolic properties of multiple compounds in the same experiment, …

• not just one each in separate experiments.


Properties Required

• The ISL must be constructed so that it can eventually function as an organ component within a larger whole organism model

• Are these properties achievable using current modeling methods?

• NO


1st Example :

• Simulated Epithelial Cell Morphogenesis In Vitro


Cell Morphogenesis In Silico

Embedded

Matrix Surface

Suspension

Overlay


Synthetic Method

• Discretize referent: into recognizable components

• Start at low level of resolution

Three Environmental Components

Generators:

Axioms

Spaces:

Hexagonal grids


Cell Component Behavior Axioms:

Otherwise, do nothing


Cell Morphogenesis In Silico


In Silico vs In Vitro


What happens if you ignore an axiom?


Behaviors Not Built Into the Axioms

Surface repairautopoiesis

Tubulogenesis


2nd Example :

• In Silico Liver


Form & Function: Levels of Organization

Lobule Secondary Unit Lobe

PV CV

PV CV

Lb

SEC.UNIT


Illustration of SinusoidsWithin Lobules


“Form” Within Lobules

Sinusoidal Segments(SSs)


“Function” Use Structured Agents


An Example of ResultsA Sucrose Outflow Profile


Question Break


Acknowledgements

• Mark Grant, Shahab Sheikh-Bahaei, Li Yan, Jon Tang, and Yu Liu – Current & Former Bioengineering Graduate Students• Members, Biosystems Group, UCSF

• Funding: NIH, CDH Resh. Fdn.

Documents

© 2006 UCSF Biosystems Group 1 EO Modeling Workshop Tucson May 16, 2006 C. Anthony Hunt UCSF Biosystems Group Designing and Building In Silico Analogues