∑k - computational & systems biology · machinery/gnm co lective dynamics gnm Å - 1 0 nm...
TRANSCRIPT
1
Cellular Simulations:Cellular Simulations:Analytic, Finite Difference, & Monte Analytic, Finite Difference, & Monte
Carlo Approaches to ReactionCarlo Approaches to Reaction--Diffusion Diffusion SystemsSystems
Joel R. Stiles, MD, PhDJoel R. Stiles, MD, PhD
NIHNIH--NSF BBSI: Simulation and Computer NSF BBSI: Simulation and Computer Visualization of Biological Systems at Multiple Visualization of Biological Systems at Multiple
ScalesScales
June 27, 2003June 27, 2003
Monte Carlo Probabilities for Monte Carlo Probabilities for UnimolecularUnimolecular TransitionsTransitions
o0
0
o0
21
)S(
)S(1
)S(
)S...()S()S( ttn
ttktp −=++=
If this reaction proceeds for any time t from an initial concentration (S0)o, the total probability (pkt) that a single molecule in the S0 state undergoes a transition is given by the fraction of(S0)o that undergoes any transition during time t:
Monte Carlo Probabilities for Monte Carlo Probabilities for UnimolecularUnimolecular TransitionsTransitions
From this, the lifetime of S0 is exponentially distributed with a mean value given by τ:
−= ∑ tk
n
it
1o0
0
exp)S(
)S(
∑=τn
ik1
1
Monte Carlo Probabilities for Monte Carlo Probabilities for UnimolecularUnimolecular TransitionsTransitions
From:
−−= ∑ tkp
n
ikt1
exp1
o0
0
o0
21
)S(
)S(1
)S(
)S...()S()S( ttn
ttktp −=++=
−= ∑ tk
n
it
1o0
0
exp)S(
)S(
and:
we obtain:
∆
−−= ∑ tkp
n
ikt1
exp1
kt
n
kin
i
nktknn
i
ktk ppk
kpp
k
kpp =⋅=⋅= ∑
∑∑ 1
11
11 ;, �
Monte Carlo Probabilities for Monte Carlo Probabilities for UnimolecularUnimolecular TransitionsTransitions
And finally:
Interactive Interactive MCellMCell Demos:Demos:
2
Monte Carlo Probabilities for Bimolecular AssociationsMonte Carlo Probabilities for Bimolecular Associations
pbt = 1 – (1 – pb)N
H
tkpn
it ∆
=ζ≅ ∑ + o
1
)A(
tkppn
itN
bH ∆
=ζ≅=−− ∑ + o
1
)A()1(1
Monte Carlo Probabilities for Bimolecular AssociationsMonte Carlo Probabilities for Bimolecular Associations
After integration, final analytic expressions for pt are:
( ) ( ) ( )[ ]( )( ) ( )[ ]( )
−⋅−−−−+
−−⋅+==
∑∑∑
+
+
qtkqbqb
qtkqbp
i
iti
t
exp2
1exp
)R(
1
)R(
AR 2
oo
2oo
oo
oo
)R)(A(4
and )RA(
where;R)((A) if
bq
b
−=
+−=≠
( ) ( )( )
( )( ) tk
tk
tk
tkp
i
i
i
it
i
t
o
o
o
2o
oo A)(1
)A(
A)(1
)A(
)R(
1
)R(
AR
∑∑
∑∑∑
+
+
+
+
+=
+==
oo )R()A( if =
or
Monte Carlo Probabilities for Bimolecular AssociationsMonte Carlo Probabilities for Bimolecular Associations
Reality check:
From these equations, the probability (pt) that a single R molecule becomes bound during an arbitrarily long interval of time (t) depends on all the k+ values, (A)o, and (R)o; pt = 0 for t = 0, and for t = ∞, pt = 1 if (A)o ≥ (R)o, or pt = (A)o/(R)o if (A)o < (R)o.
( )2
1
1111 2
;
∆π
⋅σ⋅=ΧΧ⋅⋅++Χ⋅⋅=Χ⋅⋅== +++∑∑
La
EGAnsns
n
isi
n
bib D
t
N
fkfkfkfpp �
Monte Carlo Probabilities for Bimolecular AssociationsMonte Carlo Probabilities for Bimolecular Associations
21
1 ))((2
1
∆π
= ∑ +
LETa
n
ib D
t
ANkp
Interactive Interactive MCellMCell Demos:Demos:Cells Are Complicated …Cells Are Complicated …
PSC’s Biomedical Supercomputing Initiative – An NIH Resource Center
““[[R]esultsR]esults to date show a dizzying array of signaling systems to date show a dizzying array of signaling systems acting within and between cells. … In such settings, intuition cacting within and between cells. … In such settings, intuition can an be inadequate, often giving incomplete or incorrect predictions.be inadequate, often giving incomplete or incorrect predictions. … … In the face of such complexity, computational tools must be In the face of such complexity, computational tools must be employed as a tool for understanding.” Fraser & Harland, employed as a tool for understanding.” Fraser & Harland, Cell Cell 100:41 (2000).100:41 (2000).
•• Need to accurately assess the Need to accurately assess the variabilityvariability of biological of biological systems, and their propensity to systems, and their propensity to switchswitch between between different operating modes and/or to different operating modes and/or to failfail
3
Cells Are Complicated …Cells Are Complicated …
PSC’s Biomedical Supercomputing Initiative – An NIH Resource Center
Requires: Requires: Stochastic methodsStochastic methods applied to applied to complex 3D reactioncomplex 3D reaction--diffusion diffusion andand cellscells--asas--machines machines problems problems -- largely embryonic largely embryonic due to the scope of necessary software developmentdue to the scope of necessary software development
Dramatically increasedDramatically increasedspace and time scalesspace and time scales
∆∆∆∆∆∆∆∆tt ~10~10--66 ss
MicrophysiologicalMicrophysiological (3(3--D Reaction/Diffusion) SimulationsD Reaction/Diffusion) Simulations
? How ?
QM/MMQM/MM∆∆∆∆∆∆∆∆tt ~10~10--1515 ss
Cellular StructuresMolecular LocationsMass Action Rate Constants
Molecular DiffusionMolecular MechanismsMolecular Structure
MCellMCell & & DReAMMDReAMMBuild models and Build models and use Monte Carlo SSL algorithms to couple use Monte Carlo SSL algorithms to couple
Brownian Dynamics diffusion with binding, unbinding, Brownian Dynamics diffusion with binding, unbinding, conformational changes,…conformational changes,…
Molecular & Cellular Modeling and Simulation Methods
Problem/Method
TypicalApplication
SoftwareExamples
Resolution(Scale)
SpatialRealism
StochasticRealism
TimeStep
TimeScale
Serial/Parallel
ComputerTime
Networks ofReactions/
Sets of ODEs
Metabolic orsignalingpathways
VCell, ECellGepasi
XPPAUT
N/A(cell) N/A none ms ms - hrs serial minimal
Excitation/Compartmental
Circuit
Nervesignaling
GENESISNEURONNEOSIM
µm - mm(cell -
multicell)
low -medium none ms ms - hrs usually
serialusually
low
ReactionKinetics/
Stochastic
Generegulation/
transcription
BioSpiceStochSimXPPAUT
MCell
N/A(cell) N/A high ms ms - hrs serial low
3-D ReactionDiffusion/
Finite Element
Flow models,Calcium
dynamics
VCellFIDAP
Kaskade
<µm(cell)
medium-high none µs - ms µs - sec either low - high
3-D ReactionDiffusion/
Monte Carlo
Micro-physiological
processesMCell
nm – mm(subcell -
cell)high high ps - ms µs - sec either low - high
MacromolecularMachinery/GNM
Collectivedynamics GNM Å - 100 nm
(complexes)high none N/A ps - 10 ns N/A
(analytic) minimal
Diffusion inPotential
Field/Poisson-Nernst-Planck
Electrostaticinteractions,ion channels
-----Å - nm
(membraneproteins)
high(implicitsolvent)
none N/A 10 ns parallel low -medium
MacromolecularMotions/Brownian
Dynamics (BD)
Conformationaldynamics (inflow fields)
CHARMMGROMOS
Å - nm(macro-
molecules)
high(implicitsolvent)
high 5 - 10 fs 1 - 10 ns parallel medium- high
MolecularStructure/Molecular
Dynamics (MD)
Conformationdynamics
& free energies
AMBERCHARMMGROMOS
Å(macro-
molecules)
exact(explicitsolvent)
exact 1 - 2 fs ~ ns parallel very high
MolecularStructure/Ab initio
simulations
Solution of theSchrodinger
equationGaussian98
< Å(electrons -
atoms)exact exact --- N/A parallel highest
Molecular & Cellular Modeling and Simulation MethodsMolecular & Cellular Modeling and Simulation Methods
MeshMeshGGenerationeneration
Model DesignModel Design
SimulationSimulation
AnalysisAnalysis
Various Software PackagesVarious Software Packagese.g., FormZ, XVoxTrace, NWGrid, Mesquite, LaGrit, PSC Volume Browser
DReAMMDReAMMDesign, Render & Animate MCell Models
MCellMCellGeneral Monte Carlo Simulator of Microcellular Physiology
PSC’s Biomedical Supercomputing Initiative – An NIH Resource Center
MicrophysiologicalMicrophysiological ModelingModeling
Geometry DesignGeometry Design
MeshMeshGGenerationeneration
Skipping over …Skipping over …••Brownian Dynamics Random WalkBrownian Dynamics Random Walk
(Grid(Grid--free)free)••Monte Carlo Probabilities for:Monte Carlo Probabilities for:
UnimolecularUnimolecular TransitionsTransitionsBimolecular AssociationsBimolecular Associations
••Numerical AccuracyNumerical Accuracy••RunRun--time Optimizationstime Optimizations
4
MicrophysiologicalMicrophysiological ModelingModeling
PSC’s Biomedical Supercomputing Initiative – An NIH Resource Center
MCellMCell simulations of miniature endplate currentssimulations of miniature endplate currents
Insight #1: Insight #1: JunctionalJunctional Folds Decrease Folds Decrease mepcmepc AmplitudeAmplitude
Skipping over …Skipping over …••Parameter Fitting for Parameter Fitting for NMJsNMJs::
Untreated; Untreated; AChEAChE Inhibited;Inhibited;AChEAChE Inhibited + Inhibited + AChRAChR Blockade Blockade
••Predicted Untreated Predicted Untreated mepcmepc ttrr
Testing Prediction for Untreated Testing Prediction for Untreated mepcmepc ttrr::Simultaneous Broadband EC and VC RecordingsSimultaneous Broadband EC and VC Recordings
Skipping over …Skipping over …••Quantitative Bandwidth Quantitative Bandwidth
DeterminationDetermination••Effect of EC Electrode Diameter Effect of EC Electrode Diameter
and Pressureand Pressure••mepcmepc Alignment and AveragingAlignment and Averaging
Insight #2: Insight #2: mepcmepc ttrr and and AChACh ExocytosisExocytosis Insight #3a: Temperature Sensitivity of Insight #3a: Temperature Sensitivity of mepcsmepcs is Mostly is Mostly Governed by Offsetting Effects on Channel GatingGoverned by Offsetting Effects on Channel Gating
5
Skipping over …Skipping over …••Experimental Determination ofExperimental Determination of
AChEAChE Site DensitySite Density••Experimental Determination ofExperimental Determination of
AChEAChE Turnover NumberTurnover Number
Insight #3b: Highly Nonlinear Sensitivity of Insight #3b: Highly Nonlinear Sensitivity of mepcmepc ttff to to AChEAChE ParametersParameters Insight #3a: Temperature Sensitivity of Insight #3a: Temperature Sensitivity of mepcsmepcs is Mostly is Mostly Governed by Offsetting Effects on Channel GatingGoverned by Offsetting Effects on Channel Gating
Patient followed from birth:
• Progressive weakness and impaired neuromuscular transmission without early degenerative endplate changes typically associated with SCCMS
• Prolonged, low amplitude synaptic currents at early and late stages
• Atypical, initially mild (focal) ultrastructural changes progressed over time
• Novel C-to-T substitution in exon 8 of the δδδδ subunit of AChR: serine to phenylalanine mutation in the second transmembrane domain (M2) that lines the ion channel
• AChR numbers not significantly reduced
PSC’s Biomedical Supercomputing Initiative – An NIH Resource Center
Insight #4: Synaptic Insight #4: Synaptic PathophysiologyPathophysiology in a Novel Form ofin a Novel Form ofSlow Channel Congenital Slow Channel Congenital MyasthenicMyasthenic Syndrome (SCCMS)Syndrome (SCCMS)
Normal
Focal (early)
Global(late)
Insight #4: Synaptic Insight #4: Synaptic PathophysiologyPathophysiology in a Novel Form ofin a Novel Form ofSlow Channel Congenital Slow Channel Congenital MyasthenicMyasthenic Syndrome (SCCMS)Syndrome (SCCMS)
6
Insight #4: Synaptic Insight #4: Synaptic PathophysiologyPathophysiology in a Novel Form ofin a Novel Form ofSlow Channel Congenital Slow Channel Congenital MyasthenicMyasthenic Syndrome (SCCMS)Syndrome (SCCMS)
•• Typical of SCCMS, mutant neurotransmitter receptors showed Typical of SCCMS, mutant neurotransmitter receptors showed dramatically slowed deactivation (ion channel closing rate)dramatically slowed deactivation (ion channel closing rate)
•• However, However, simulations simulations of synaptic signalsof synaptic signals predictedpredicted an an additional noveladditional novel slowing of receptor activation (ion channel slowing of receptor activation (ion channel opening rateopening rate))
•• Based on model predictions, opening rate was measured and Based on model predictions, opening rate was measured and found to be decreasedfound to be decreased -- likely explains unique early course of likely explains unique early course of disease in this patientdisease in this patient
PSC’s Biomedical Supercomputing Initiative – An NIH Resource Center
200open
AChRs1 ms
Insight #5: Insight #5: SpatioSpatio--Temporal Correlations and Synaptic NoiseTemporal Correlations and Synaptic Noise
averaged results from averaged results from MCellMCell simulations (3000 runs)simulations (3000 runs)
SpatioSpatio--Temporal Correlations and Synaptic NoiseTemporal Correlations and Synaptic Noise
Stochastic Simulation Algorithm Stochastic Simulation Algorithm ModelModel
Insight #5: Insight #5: SpatioSpatio--Temporal Correlations and Synaptic NoiseTemporal Correlations and Synaptic Noise Insight #5: Insight #5: SpatioSpatio--Temporal Correlations and Synaptic NoiseTemporal Correlations and Synaptic Noise
7
Insight #5: Insight #5: SpatioSpatio--Temporal Correlations and Synaptic NoiseTemporal Correlations and Synaptic Noise Insight #5: Insight #5: SpatioSpatio--Temporal Correlations and Synaptic NoiseTemporal Correlations and Synaptic Noise
PresynapticPresynaptic calcium dynamics & neurotransmitter releasecalcium dynamics & neurotransmitter release
NormalNormal Developing/RegeneratingDeveloping/Regenerating
PresynapticPresynaptic calcium dynamics & neurotransmitter releasecalcium dynamics & neurotransmitter release
NormalNormal
PresynapticPresynaptic calcium dynamics & neurotransmitter releasecalcium dynamics & neurotransmitter release
8
Developing/RegeneratingDeveloping/Regenerating
PresynapticPresynaptic calcium dynamics & neurotransmitter releasecalcium dynamics & neurotransmitter release
33--D reconstruction of D reconstruction of vertebrate neuromuscular vertebrate neuromuscular
junctionjunction
Synaptic topology & current variabilitySynaptic topology & current variability
PSC’s Biomedical Supercomputing Initiative – An NIH Resource Center
Synaptic topology & Synaptic topology & current variabilitycurrent variability
Synaptic topology & current variabilitySynaptic topology & current variability Synaptic topology & current variability Synaptic topology & current variability –– synaptic synaptic AChEAChE distributiondistribution
9
PSC’s Biomedical Supercomputing Initiative – An NIH Resource Center
Usage Scenarios:
Expanding User Group & Range of ApplicationsExpanding User Group & Range of Applications
Serial or Parallel - Look & See, Parameter Fitting, Parameter SweepMainframes Workstations Parallel Grid
Workshops1997 – Cornell Theory Center, NSF support (MCell test group)
1999 – SDSC, NSF support (MCell test group)
2001 – PSC, NCRR support (open, MCell & DReAMM)
2003 – PSC, (Drosophila neuromuscular research, MCell & DReAMM)
Outside Users:Outside Users:Registration & Download Website Maintained at PSC – UNIX, PC, Mac
>300 Registered>300 Registered
WebWeb--based Instructionbased Instructionwww.mcell.psc.eduwww.mcell.psc.edu
www.mcell.psc.edu/DReAMMwww.mcell.psc.edu/DReAMM
Developers & CollaboratorsDevelopers & Collaborators
Joel R. Stiles (PSC, CMU, Pitt)Joel R. Stiles (PSC, CMU, Pitt)Thomas M. Thomas M. BartolBartol (Computational Neurobiology Lab (Computational Neurobiology Lab
of Terrence J. of Terrence J. SejnowskiSejnowski, The Salk Institute), The Salk Institute)
Edwin E. Edwin E. SalpeterSalpeter (Physics and Astronomy, Cornell)(Physics and Astronomy, Cornell)
MCellMCell and and DReAMMDReAMM::
PSC’s Biomedical Supercomputing Initiative – An NIH Resource Center
NSF ITR NSF ITR -- Virtual Instruments: Scalable Virtual Instruments: Scalable Software Instruments for the Grid:Software Instruments for the Grid:
Berman, F. (PI, UCSD)Berman, F. (PI, UCSD)Casanova, H. (UCSD)Casanova, H. (UCSD)DongarraDongarra, J. (U. Tenn.) , J. (U. Tenn.) EllismanEllisman, M. (UCSD), M. (UCSD)WolskiWolski, R. (UCSB), R. (UCSB)
MCellMCell Developers &Developers &
AnglisterAnglister, L., L. (Hebrew U.)(Hebrew U.)DeerinckDeerinck, T., , T., EllismanEllisman, M., M. (UCSD)(UCSD)Ford, W.Ford, W. (U. Pittsburgh)(U. Pittsburgh)Gomez, C.Gomez, C. (U. Minnesota)(U. Minnesota)KovyazinaKovyazina, I., I. ((KazanKazan U.)U.)PattilloPattillo, J., , J., MerineyMeriney, S., S. (U. Pittsburgh)(U. Pittsburgh)SalpeterSalpeter, M., M. (Cornell)(Cornell)TemizTemiz, N.A., , N.A., BaharBahar, I., I. (U. Pittsburgh)(U. Pittsburgh)Van Van HeldenHelden., D.., D. (U. Newcastle)(U. Newcastle)
MicrophysiologyMicrophysiology::
NIHNIH--NSF BBSI: Simulation and Computer Visualization of Biological SyNSF BBSI: Simulation and Computer Visualization of Biological Systems at stems at Multiple Scales Multiple Scales –– Summer Institute beginning May 2003; Pitt, PSC, Duquesne, CMUSummer Institute beginning May 2003; Pitt, PSC, Duquesne, CMU
“Computational biology today is in a state analogous to weather prediction 25 years ago, when it was not very reliable or highlyrespected. … Now we routinely rely on accurate global weather forecasting. Similarly, we foresee a new age of cell biology, when computational modeling plays an integral role in our understanding of the molecular basis of cell physiology and human disease. The impact of computational cell biology on biomedicine will be just as great as the impact of modern weather prediction on the transportation industry, the military, and civil defense.”
NARRC Working Group on Computational Cell Biology, September 10, 2002, Farmington, ConnecticutGary Borisy, Arey Professor of Cell and Molecular Biology, Northwestern; President, American Society for Cell BiologyJames Cassatt, Director of Cell and Molecular Biology and Acting Director Center for Computational Biology, NIGMS
Ann Cowan, Assistant Professor of Biochemistry and Deputy director, Center for Biomedical Imaging Technology, UConnHC.Amarnath Gupta, Associate Research Scientist, San Diego Supercomputer Center
Chris Hogue, Asst. Professor of Biochemistry and Director of Bioinformatics, U. Toronto; Director of the BIND ProjectMike Hucka, Researcher, Control and Dynamical Systems, CalTech; SBW/SBML
Ravi Iyengar, Interim Dean for Research, Rosenstiel Professor and Chair of Pharmacology and Biochemistry, Mt. Sinai Medical SchoolLeslie Loew, Professor of Physiology, UConnHC; Director, National Resource for Cell Analysis and Modeling
Robert Phair, President, Bioinformatics Services, Inc.Tom Pollard, Higgins Professor of Cell, Molecular and Developmental Biology, Yale
Brian Ray, Editor, Science STKERalph Roskies, Professor of Physics, U. Pittsburgh and Scientific Director, Pittsburgh Supercomputing Center
Dong-Guk Shin, Professor of Computer Science and Engineering, UConnJoel Stiles, Senior Scientist, Pittsburgh Supercomputing Center; MCELL
John Tyson, University Distinguished Professor of Molecular Cell Biology, Virginia TechRai Winslow, Professor of Biomedical Engineering and Director Center for Computational Medicine and Biology, Johns Hopkins