Deciphering molecular interactions using HPC simulations: getting new therapeutic targets

Transcription

1 IX ème journée 2 juillet 2014 Deciphering molecular interactions using HPC simulations: getting new therapeutic targets Pr Manuel Dauchez Laboratoire SirMa CNRS UMR 7369 MEDyC Maison de la Simulation de Champagne-Ardenne Université de Reims-Champagne-Ardenne

2 Biological context Applications & Biological Experiments Dedicated Developments HPC Bioinformatics, Structural Biology in silico DataBanks Methodologies

3 Why to use simulations explain experiment, provoke experiment, replace experiment, aid in establishing intellectual property 3D structures structural adaptability functional prediction Structure/function/dynamics relationships «in silico veritas»

4 Multiscale problem Bottom-up approach >>>> <<<< Upside-down approach

5 Multiscale problem m m 10-9 m 10-8 m 10-7 m 10-6 m 10-2 m HPC, a Computational Microscope: a way to study molecular machines of living cell!

6 Multiphysics problem 1-2 peptides 1 association <10 2 atoms «10 ns» 1-2 proteins 1 complex 10 4 atoms 100 ns 5-10 proteins 2-3 complex 10 5 atoms 1 µs >1000 proteins >100 complex >10 6 atoms >1 ms Quantum Dynamics Molecular Dynamics Normal Modes Coarse Grained or Langevin Dynamics Mesoscopic Dynamics

7 Methodology to visualise to model to simulate «Maison de la Simulation de Champagne-Ardenne»

8 Empirical FF Empirical Force Fields & programs (Gromos, NAMD, Amber, Charmm...) F=ma protein Metals: Zn, Ca Na + r i (t) r i (t + dt) s 7-10 orders of magnitude N atoms, N 2 Memory, N 3 time CPU & GPU Local motion of peptides and proteins Coarse-grained OPEP, Martini Longer simulations and larger systems, but simplified! Cl - Water

9 Others approaches Normal Modes Analysis Analytical solving of the equations of motion eigenvalues = frequencies eigenvectors = modes Global motions of proteins Minimum CPU & GPU Functional motions Docking + - Interactions ligands/proteins & proteins/proteins Surfaces + Electrostatic + Hydrophobic adjustements

10 Biological context Polysacharrides E C M Glycosaminoglycannes Fibrous proteins (Elastin / collagen) Laminin, Fibronectin water Mecanical properties [ bones, elasticity of skin ] Biological properties [ Cellular signals, inflammation ] Thrombospondin, LRP Degradation MMPs Regulation with TIMPs Extracellular Matrix ECM Elastin peptides/ebp

11 Thrombospondin TSP1

12 EGF- like Repeats General Properties Régi ons gl obul ai r es NH2 RGD A Multimeric extracellular Calcium matrix protein Ion Between 1154 and 737 Mannose amino acids and per Fucose monomer (human TSP) N- Type Stalk Repeat Multiple domains housing multiple functions C- Type Stalk Repeat Involved in inflammation, proliferation, angiogenesis, apoptosis, Globe tumorigenesis, PSACH... dom ai ne de de t ype I Disulphide Bond connexi on Répét i t i ons Répét i t i ons de t ype III RGD A Régi ons gl obul ai res CO O H dom ai ne Pr ocol l agène Répét i t i ons de t ype II RGD A

13 Simulation of TSP1 Molecular Dynamics domain by domain Behaviour of «atomic parts» Normal Mode Analysis Detection of interactions using simulations New candidates Validation using HPC simulations Experimental validation

14

15

16 Simulation of TSP1

17 Interaction of TSP1/CD47

18 Design of a peptide

19 Simulation of the peptide

20 Biological effects

21 Biological effects use of GPU : hours minutes

22 Inverse docking: how to get new molecules!

23 Problematic & scientific context Fibrillar proteins Elastin, collagen, Elastases Collagenases Which peptide/drug is able to interact with which ECM protein Active peptides Matrikins Matricryptins Drugs Chemical ligands HPC Inverse Docking 1 2 developments for ligands developments for peptides

24 Inverse docking procedure molecule of interest target molecule One ligand Inverse Virtual Ligand Screening Protein structures databases CPU 1 CPU 2 CPU 3 CPU k

25 Arbitrary cutting in multiples boxes druggables cavities searchs Fpocket 1 1. Fpocket : V. Le Guilloux, P. Schmidtke, and P. Tuffery, MTi, Paris. blind docking

26 Task scheduling and Distribution AutoDock 4.2

27 User Needs Faster Science, less power consumption, better results Application and Performance Experts PhD in different scientific domains Focussing on Performance Bull Software Stack and Tools Dedicated Hardware POC and Experiments (CPU, Accelerators, )

28 First application of «inverse docking»

29 First application of «inverse docking» EGFEPG GVAPGV GVGVAP APGVGV LGTIPG PGAIPG VAPGVG VGVAPG PGVGVA PGAYPG

30 Second application of «inverse docking»

31 Second application of «inverse docking»

32 Second application of «inverse docking»

33 Biological associated effects Témoin TIMP-1 M1 M2 M3

34 Quantum docking: how to get new interactions!

35 «Quantum Docking» Using Quantum Mechanics to evaluate affinity score between ligand and receptor. HΨ=EΨ Very HUGE calculations New quantum methodologies called «linear» possibility to calculate very large systems at electronic levels for instance in active sites in Proteins HPC NEEDED! Access to polarization effects, water behaviour

36 Some shortcuts!

37 FUTURE!

38 New TOOLS Big Data to visualise to model New TOOLS Big Data before after to simulate Exascale!!