multi scale simulations of whole
play

Multi-scale simulations of whole Tyler Earnest / PI: Zan - PowerPoint PPT Presentation

Multi-scale simulations of whole Tyler Earnest / PI: Zan Luthey-Schulten (UIUC) NCSA Blue Waters Symposium for Petascale Science and yeast cells Beyond: June 6, 2018 Chemical kinetic simulations Earnest, , ZLS Rep. Prog. Phys (2018)


  1. Multi-scale simulations of whole Tyler Earnest / PI: Zan Luthey-Schulten (UIUC) NCSA Blue Waters Symposium for Petascale Science and yeast cells Beyond: June 6, 2018

  2. Chemical kinetic simulations Earnest, …, ZLS Rep. Prog. Phys (2018) 10.1088/1361-6633/aaae2c

  3. Chemical kinetic simulations c ( t ) Earnest, …, ZLS Rep. Prog. Phys (2018) 10.1088/1361-6633/aaae2c

  4. Chemical kinetic simulations c ( t ) c ( x , t ) Earnest, …, ZLS Rep. Prog. Phys (2018) 10.1088/1361-6633/aaae2c

  5. Chemical kinetic simulations P ( n , t ) c ( t ) c ( x , t ) Earnest, …, ZLS Rep. Prog. Phys (2018) 10.1088/1361-6633/aaae2c

  6. Chemical kinetic simulations P ( n , t ) c ( t ) P ( n , x , t ) c ( x , t ) Earnest, …, ZLS Rep. Prog. Phys (2018) 10.1088/1361-6633/aaae2c

  7. CME/ODE — Well mixed chemical kinetics

  8. Saccharomyces cerevisiae — Characteristic scales Di ff . coe ff . Conc. [µM] Di ff . time [s] [µm 2 /s] Sugar ~1,000 ~200 0.013 Protein ~1 ~5 0.53 4 µm Gene 10 -4 10 -5 10 5

  9. Saccharomyces cerevisiae — Characteristic scales Di ff . coe ff . Conc. [µM] Di ff . time [s] [µm 2 /s] Sugar ~1,000 ~200 0.013 Protein ~1 ~5 0.53 4 µm Gene 10 -4 10 -5 10 5 Concentration/time scales do not overlap!

  10. Yeast galactose switch Ramsey et al. Nat Genet. 2006 10.1038/ng1869

  11. Yeast galactose switch 10 7 reactions per second over 6 hours Ramsey et al. Nat Genet. 2006 10.1038/ng1869

  12. Yeast galactose switch 10 7 reactions per second Switch induction result of but over 6 hours by single molecule event Ramsey et al. Nat Genet. 2006 10.1038/ng1869

  13. Hybrid CME/ODE simulations Bianchi, …, TME, …, ZLS. IET Syst. Biol. 2018 10.1049/iet-syb.2017.0070

  14. Hybrid CME/ODE simulations Chemical master equation: Gillespie algorithm Bianchi, …, TME, …, ZLS. IET Syst. Biol. 2018 10.1049/iet-syb.2017.0070

  15. Hybrid CME/ODE simulations Chemical rate equation: LSODA Chemical master equation: Gillespie algorithm Bianchi, …, TME, …, ZLS. IET Syst. Biol. 2018 10.1049/iet-syb.2017.0070

  16. Hybrid CME/ODE simulations Communication Timestep τ Chemical CME rate equation: CME updates ODE species counts LSODA ODE updates CME propensities ODE time ⎧ ⎬ ⎫ ODE Adaptive Timestep ∆ t Chemical master equation: Gillespie algorithm Bianchi, …, TME, …, ZLS. IET Syst. Biol. 2018 10.1049/iet-syb.2017.0070

  17. Results: 40x speed up with limited impact on accuracy Wall time to simulate 750 min External galactose 0.055 mM 2.0 mM CME (hr) 2.1 47.4 Hybrid (10 s) 0.4 (5.2) 1.1 (43.1) Hybrid (1s) 0.8 (2.6) 1.8 (26.3)

  18. RDME/ODE — Spatial resolution

  19. Reaction-Diffusion Master Equation

  20. Reaction-Diffusion Master Equation

  21. Reaction-Diffusion Master Equation Discretize to lattice

  22. Reaction-Diffusion Master Equation Discretize to lattice Particles: React within subvolumes Diffuse between subvolumes

  23. Reaction-Diffusion Master Equation Discretize to lattice Particles: React within subvolumes Diffuse between subvolumes System state described by particle count of each type at each subvolume

  24. 
 
 Whole-cell modeling with Lattice Microbes Designed for CUDA Multiple levels of parallelism GPGPU 
 Roberts et al. IPDPS (2009) 10.1109/ipdps.2009.5160930 Multi-GPU 
 Hallock et al. Parallel Comput . 2014 10.1016/j.parco.2014.03.009 MPI 
 (w.i.p.) Extensible through Python Peterson et al. PyHPC 2013 10.13140/2.1.3207.7440 
 ⟹ Hybrid solvers Roberts et al. J. Comput. Chem. 2013

  25. Yeast galactose switch — Initial geometry Cryo-electron tomography 2.04 × 1.91 × 0.242 μ m 3 section of yeast cell 2% of total volume Use tomogram as template to build whole cell geometry TME, …, E. Villa, ZLS. 10.1021/acs.jpcb.7b00672

  26. Use tomogram as template for cell geometry

  27. Use tomogram as template for cell geometry

  28. Use tomogram as template for cell geometry

  29. Use tomogram as template for cell geometry

  30. G80 is sequestered from nucleus

  31. New transporters are localized near nucleus

  32. New transporters are localized near nucleus

  33. Outlook Demonstrated CME ⟷ ODE RDME ⟷ ODE Planning RDME ⟷ PDE In progress RDME ⟷ Brownian dynamics

  34. Outlook Demonstrated CME ⟷ ODE RDME ⟷ ODE Planning RDME ⟷ PDE In progress RDME ⟷ Brownian dynamics

  35. Outlook Goal Whole-cell physico-chemical model of minimal cell JCVI-syn3A 1 400 nm diameter 453 protein coding genes 2 hr doubling time Multi-scale modeling ODE —Metabolism 2 RDME — Gene expression 3 BD —Ribosomes, DNA, cytoskeleton 1 Huchison et al. Science 2016 10.1126/science.aad6253 2 Breuer, TME, …, ZLS 2018 submitted 3 TME, …, ZLS. BPJ 2016 10.1016/j.bpj.2015.07.030

  36. Acknowledgements David Bianchi Mike Hallock Joe Peterson Zan Luthey-Schulten

  37. Acknowledgements David Bianchi Mike Hallock Joe Peterson Zan Luthey-Schulten Lattice Microbes: http://goo.gl/akkSyg

Recommend


More recommend