biological science libraries - scripps research institute science libraries - an approach to...

Biological Science LibrariesBiological Science Libraries- An approach to software- An approach to softwarecomponents for structuralcomponents for structural

bioinformaticsbioinformatics

Michel F. Sanner

The Scripps Research InstituteLa Jolla, California

The Molecular Graphics Laboratory

SciPy’02

CalTech Pasadena CA. Sept. 2002

SciPy’02, © M.Sanner, TSRI

SCHEDULESCHEDULE

o Component-based software? Background and motivation? Re-usable components

o From building blocks to applications? PMV: a molecular visualization program? ADT: setting up and running AutoDock? PVV: interactive volume rendering? ViPEr: a visual programming environment

o Conclusion

SciPy’02, © M.Sanner, TSRI

Protein-Protein-LigandLigand

Complex AssembliesComplex AssembliesProtein-ProteinProtein-Protein

Molecular SurfacesMolecular SurfacesAutoDockAutoDock

SurfDockSurfDock

SymGenSymGen

MSMSMSMS

HARMONYHARMONY

SymSearchSymSearch

SciPy’02, © M.Sanner, TSRI

Tangible ModelsTangible Models

SciPy’02, © M.Sanner, TSRI

ARToolkitARToolkit 3D tracking 3D tracking

SciPy’02, © M.Sanner, TSRI

Augmented realityAugmented reality

SciPy’02, © M.Sanner, TSRI

VisualizationVisualization

DockingDockingMethodsMethods

FoldingFolding

ProteinProteinEngineeringEngineering

SequenceSequenceAnalysisAnalysis

ModelingModeling

MM - MDMM - MD

Ab InitioAb InitioMethodsMethods

ElectrostaticsElectrostaticsCalculationsCalculations

MolecularMolecularSurfacesSurfaces

Etc ...Etc ...

The challengeThe challenge

SciPy’02, © M.Sanner, TSRI

VisualizationVisualization

DockingDockingMethodsMethods

FoldingFolding

ProteinProteinEngineeringEngineering

SequenceSequenceAnalysisAnalysis

ModelingModeling

MM - MDMM - MD

Ab InitioAb InitioMethodsMethods

ElectrostaticsElectrostaticsCalculationsCalculations

MolecularMolecularSurfacesSurfaces

Etc ...Etc ...

“Traditional” solution“Traditional” solution

N2 interfaceslow interoperability

No code Reuse

SciPy’02, © M.Sanner, TSRI

High level language as a scripting environmentHigh level language as a scripting environment

MolecularSurfaces

Molecular

MolecularSurfaces

Surfaces

Molecules

Molecules

Molecules

DataBase

DataDataBaseBaseElectrostatics

Electrostatics

Electrostatics

DelaunayDelaunayDelaunay

Homology

Homology

HomologyCSGCSGCSG

3DViewer

3D3DViewerViewer

MM-MDMM-MDMM-MD

NewMethod

NewNewMethodMethod

Your

MethodYourYour

MethodMethod

Python to the rescuePython to the rescue

Interactive

Dynamic

Platformindependent

SciPy’02, © M.Sanner, TSRI

Writing an applicationWriting an application

o High level codingo Code re-useo modularityo ExtensibleMolecules

Molecules

Molecules3DViewer

3D3DViewer

Viewer

MolecularSurfaces

Molecular

MolecularSurfaces

Surfaces

Electrostatics

Electrostatics

Electrostatics

SciPy’02, © M.Sanner, TSRI

PYTHON - FeaturesPYTHON - Features

o Interpreted, High level, Object-Orientedo Flexible and Extensibleo Introspection, self-documentingo Platform Independento Open Sourceo Rapidly gaining acceptance

SciPy’02, © M.Sanner, TSRI

Why Python ?Why Python ?

o Advanced data structures (> Tcl, Perl)o Object-Oriented AND scripting (> Java)o “Batteries included !”

? Powerfull data-parallel Numeric arrays (> Tcl)o Readbility and modularity (> Perl)o High-level (> C, C++, Fortan)o Platform independence (> C++, Java)o Efficient enough for many things !o Automated tools for wrapping C, C++, Fortran

SciPy’02, © M.Sanner, TSRI

Re-usable componentsRe-usable componentso MolKit:

? read/write/represent/manipulate moleculeso PyBabel:

? Atom hybridization, charges, hydrogen atomso DejaVu:

? General purpose 3D geometry viewero ViewerFramework:

? Visualization application templateo Mslib, PyBabel, PyMead, SFF, Isocontour, ...

(Sophie I. Coon, Michel F. Sanner and Art J. Olson, Re-usable components for structural bioinformatic, 9th Python Conference 2001)

SciPy’02, © M.Sanner, TSRI

from MolKit.pdbParser import PdbParserparser = PdbParser(‘1crn.pdb’)mols = parser.parse( )

from MolKit.pdbParser import PdbParserparser = PdbParser(‘1crn.pdb’)mols = parser.parse( )

PDBMol2PQR...

Parser

MoleculeSet

PDB parser MOL2 parserMolecule

Chain

Residue

Atom

Molecule

Chain

Residue

Atom

Molecule

Residue

Atom

Molecule

Atom

MolKitMolKit MolKit

SciPy’02, © M.Sanner, TSRI

ExamplesExamples>>> from MolKit import Read>>> molecules = Read(‘./1crn.pdb’) # Read returns a ProteinSet>>> mol = molecules[0]>>> print mol.chains.residues.name>>> print mol.chains.residues.atoms[20:85].full_name()>>> from MolKit.molecule import Atom>>> allAtoms = mol. findType(Atom)>>> set1 = allAtoms.get(lambda x: x.temperatureFactor >20)>>> allResidues = allAtoms.parent.uniq()>>> import Numeric>>> for r in allResidues:. . . coords = r.atoms.coords. . . r.geomCenter = Numeric.sum(coords) / len(coords)

MolKit

SciPy’02, © M.Sanner, TSRI

PyBabelPyBabel

o From Babel v1.6 (mostly)? Atom types and bond order detection? Aromaticity detection? Adding Hydrogen atoms? Computing Gasteiger charges

PyBabel

SciPy’02, © M.Sanner, TSRI

TkinterTkinterNumericNumeric

PyOpenGLPyOpenGL

DejaVuDejaVu

from DejaVu import Viewervi = Viewer( )from DejaVu.Spheres import Spherescenters =[[0,0,0],[3,0,0],[0,3,0]] s = Spheres(‘sph’, centers = centers)s.Set(quality=10)vi.AddObject(s)

from DejaVu import Viewervi = Viewer( )from DejaVu.Spheres import Spherescenters =[[0,0,0],[3,0,0],[0,3,0]] s = Spheres(‘sph’, centers = centers)s.Set(quality=10)vi.AddObject(s)

DejaVuDejaVu DejaVu

SciPy’02, © M.Sanner, TSRI

FeaturesFeatureso OpenGL Lighting and Material modelo Material editor, transparencyo Arbitrary clipping planeso Multiple light sourceso DepthCueing (fog), global anti-aliasingo glSscissors/magic lenso Object hierarchyo Instanceso Multi-level pickingo Extensible set of geometries

DejaVu

SciPy’02, © M.Sanner, TSRI

DejaVu DejaVu and and MolKitMolKit

from MolKit import Readmolecules = Read(‘./1crn.pdb’)coords = molecules.chains.residues.atoms.coordsradii = mol.defaultRadii()from DejaVu import Viewervi = Viewer() # create a viewerfrom DejaVu.Spheres import Spheress=Spheres(‘sph’, centers=coords, radii=radii,quality=10)vi.AddObject(s) # display atomic spheres

from MolKit import Readmolecules = Read(‘./1crn.pdb’)coords = molecules.chains.residues.atoms.coordsradii = mol.defaultRadii()from DejaVu import Viewervi = Viewer() # create a viewerfrom DejaVu.Spheres import Spheress=Spheres(‘sph’, centers=coords, radii=radii,quality=10)vi.AddObject(s) # display atomic spheres

SciPy’02, © M.Sanner, TSRI

NumericNumericPyOpenGLPyOpenGL TkinterTkinter

DejaVuDejaVu

IdleIdle

ViewerFrameworkViewerFramework

ViewerFrameworkViewerFramework ViewerFramework

SciPy’02, © M.Sanner, TSRI

NumericNumeric

PyOpenGLPyOpenGL TkinterTkinter

DejaVuDejaVuIdleIdle

ViewerFrameworkViewerFramework

MslibMslib

PmvPmv

MolKitMolKit

PMVPMV: From Building Blocks to applications: From Building Blocks to applications

SciPy’02, © M.Sanner, TSRI

PMV CapabilitiesPMV Capabilities

o Display/compute:? Lines, CPK, Stick and Balls, Surfaces,

Splines, Ribbons, H-bonds, bond-order,Gasteiger and Kollmann charges

o Editing:? deleting atoms, adding hydrogens

o Electrostatic potential? MEAD*, APBS*

o Molecular Dynamics:? Amber*

SciPy’02, © M.Sanner, TSRI

PMVPMV: Architecture: Architecture

Numeric PyOpenGL TkinterDejaVu

ViewerFramework

IdleMolKitMslib

MsmsCommands

SciPy’02, © M.Sanner, TSRI

Automated docking of a flexible ligand to macromolecules using affinity grids

AutoDockAutoDock fundamentals fundamentals

SciPy’02, © M.Sanner, TSRI

AutoDock ToolKit AutoDock ToolKit ((ADTADT))

o AutoTors : ligand preparation

o AutoGpf : grid definition

o AutoDpf : docking parameters definition

o AutoStart : job launching and monitoring

o AutoAnalyze : docking results analysis

AutoDockTools

SciPy’02, © M.Sanner, TSRI

ADT: extending PMVADT: extending PMV

SciPy’02, © M.Sanner, TSRI

ADTADT: Architecture: Architecture

Numeric PyOpenGL TkinterDejaVu

ViewerFramework

IdleMolKitMslib

MsmsCommands

AutoTors, AutoGpfAutoDpf, AutoLaunch

AutoAnalyze

SciPy’02, © M.Sanner, TSRI

Interactive Volume RenderingInteractive Volume Rendering

o VolumePro™ 500 Key Features? Renders a 2563 volume up to 30 frames per second? 256MB of volume memory? Real-time parameter changes? Full phong based lighting model on per sample basis? (8, 8, 8, 12) bit voxels? One oblique cut plane with thickness? Six box cropping planes

? 4k color lookup table with high precision alpha blending

? Support for anisotropic and gantry tilted volume data

SciPy’02, © M.Sanner, TSRI

VolPro VolPro and and DejaVuDejaVu

o VLI? C++ library driving the VolPro card? Wrapped for Python using SWIG

o VLIGeom? Subclasses DejaVu.Geom? Ties the Geometry transformation to

the VLI Camera position? Display the result of a volume rendering

operation in the DejaVu OpenGL context

VLI

SciPy’02, © M.Sanner, TSRI

VV componentVV component VV

Set of ViewerFramework commands exposing the VLI API

SciPy’02, © M.Sanner, TSRI

Large Large DatasetDataset: Mesh: Mesho Mesh Toolkit

? PyMesh: wrapped for Python using SWIGo Zoom command

Mesh

SciPy’02, © M.Sanner, TSRI

Software Volume RenderingSoftware Volume Rendering

OmniORBo UT Austin volserv.idlo OmniORB

? C++ ORB with Python mappingsUT Austin

VolRenCORBA

SciPy’02, © M.Sanner, TSRI

PVVPVV: Architecture: Architecture

Numeric PyOpenGL TkinterDejaVu

ViewerFramework

IdleMolKitVLI

VolView

ViPErViPEr: a visual programming: a visual programming

o Basics, Architecture, Librarieso Nodes, Execution Flowo Demos

(Michel F. Sanner, Daniel Stoffler and Arthur J. Olson, ViPEr, a VisualProgramming Environment for Python, 10th Python Conference 2002)

SciPy’02, © M.Sanner, TSRI

AVS-5: AVS-5: Advanced Visualization System Inc.Advanced Visualization System Inc.

Enabling scientists (non-programmers)to build computational networksEnabling scientists (non-programmers)to build computational networks

SciPy’02, © M.Sanner, TSRI

ViPErViPEr: visual programming: visual programming

SciPy’02, © M.Sanner, TSRI

NetworEditor

ViPEr

Architecture OverviewArchitecture Overview

SciPy’02, © M.Sanner, TSRI

Standard 3D Viewer Misc.MolKitSymServerImageLibCCP4…

DejaVu

MolKitsymservPIL

ViPER Libraries Python Packages

Libraries / code reuseLibraries / code reuse

SciPy’02, © M.Sanner, TSRI

Why Python worked !Why Python worked !

o Promotes better coding:? readability? high level? modular

o Software “Lego”

SciPy’02, © M.Sanner, TSRI

What could be betterWhat could be better

o Handling of platform dependent andplatform independent code

o Version managmento Distribution of sets of compatible

modules

SciPy’02, © M.Sanner, TSRI

ConclusionConclusion

o Software components used in:? PMV: flexible and extensible molecular

visualization environment? ADT: AutoDock front/back end? ViPEr: General Purpose Visual

Programming Environmento Python promotes modularityo Code reuse

SciPy’02, © M.Sanner, TSRI

AcknowledgmentsAcknowledgments

o Molecular Graphics Lab. at TSRIo NBCR, NIH, NSFo Swiss National Science Foundationo Daniel Stoffler (SymServer, AUM)o Sophie Coon, Ruth Huey

Available at:http://www.scripps.edu/~sanner/python