Transcript 16-Processes in Space

Stochastic, Spatial and Concurrent
Biological Processes Modeling
Yifei Bao, Eduardo Bonelli, Philippe Bidinger,
Justin Sousa, Vishakha Sharma
Advisor: Adriana Compagnoni
Department of Computer Science
Joint work with Libera’s lab and Sukhishvili’s lab
from Department of CCBBME
Objective
• Construct a language to model and
simulate biological processes.
• Apply it for the modeling of a drug
delivery nano-system.
Outline
•
•
•
•
•
Motivating example: Bio Film System
Survey for Existing Modeling Techniques
Our Contribution: A Simulation Language
Ongoing and Future Work
Project Demo
Drug Delivery System
• Biofilms are loaded with antibiotics and
they are used to coat medical implants.
• When the pH changes due to infection,
the Biofilm releases molecules of
antibiotics.
Sequential release of bioactive molecules from
layer-by-layer films
Bio Film System
increasing pH
basic/neutral
fast release of
capsule cargo
3.2
μm
3.2
μm
Data from Prof. Sukhishvili’s Lab
Relationship
between release
of drug molecules
and PH with
respect to time.
Computational Model
• Motivation:
– Wet lab experiments are costly
– Some data are difficult to observe (local pH)
• Predict interactions between species
 Bacteria
 Drug Molecule
• Predict local PH
• Visualization of Bio system
SPIM
• Concurrent communicating processes
– Processes evolve concurrently
– Synchronize by message passing
• Successfully used for modeling biological systems
– Process = Molecule (with state)
– Synchronization = Reaction
•Existing implementation
• Simulation and visualization
• 4000 lines of ML (Ocaml, F#) code
SPiM Model
SPiM not suitable for Bio Film
example
• SPiM assumes reactions occur in
homogeneous mixture
• Not applicable to Bio Film example
(antibiotic stored in film – not in solution)
Spatial modeling is needed
• Reaction distance: only molecules close
enough can react.
• Reaction boundary: the movements and
reactions should occur in specific areas.
• Shape of Binding Sites : only matching shapes
can bind.
Existing modeling methods
• Lack spatial attributes: ODEs, SPiM , Kappa,
Petri Nets.
• Limited notion of space: BioAmbinet, BioPepa,
StochSim.
• Lack stochasticity: SpacePi.
• Very ad hoc models.
Our Contribution
• A language for the simulation of stochastic
biological processes with spatial
information
– An extension of the SPIM language
– Language definition and implementation
• Model of the Biofilm system
SPIM
• SPiM Assumption: all molecules (processes) are
assumed to be uniformly distributed in space
• Interactions scheduled randomly based on
concentrations and reaction rates
– Informally: interaction involving higher
concentrations and rates are more likely to occur
Gillespie algorithm
Spatial Features
• Process state includes spatial information
– Each process has a position and three vectors that
define its local system of coordinates
• This state can be modified by application of affine
maps (translation, rotation..)
– Simulation of movement (translation, rotations)
• Interactions may be conditioned by the distance
between two molecules
Spatial Features
Implementation
•Based on SPIM Interpreter
•Update of parser, type checker
•Simulation algorithm (scheduler)
•Graphical output
•Basic geometric computation (affine map
application, distance, rotation..)
Ongoing Work: Validation
• We need to validate:
1) Language design (expressivity)
2) Correctness of simulation algorithm
3) Performance
4) Biofilm model
• Involve interaction with the bio-chemistry team (esp.
for 2 and 4)
– e.g. actual physical data
Ongoing Work: Shapes
• Enrich the language to target a wider class of
systems
– Processes are modeled as immaterial points
– But physical objects have a shape
• Add shape information to processes in order to
model
– Boundaries (material that can't be crossed)
– More complex interaction patterns based on
the shape and orientation of a molecules
• Apply our technique to Wireless Communication
Demo
Her2 Signaling Pathways
09/02/10