Integrative Models of Molecular Machines: ChimeraX Demo

Tom Goddard
November 8, 2016
NIH site-visit

This ChimeraX demonstration exploring nuclear pore architecture will highlight recent developments related to Technology Research and Development (TR&D) projects 1 and 2, and two driving biological projects (DBP 1 and 3) and a collaborative project (CSP 1) described in the grant proposal.

Collaborations highlighted in this demo

• DBP 1: Integrative modeling, Andrej Sali, UCSF, and Helen Berman, Rutgers University.
• DBP 3: Data archiving formats and visualization, Steve Burley, RCSB.
• Collaboration 1: Visualizing and analyzing crosslinks, Martin Beck, EMBL Heidelberg.

Technologies highlighted in this demo

• Integrated analysis of diverse data used to build molecular models.
• Advanced graphics for visualizing very large molecular models (millions of atom).
• Interactive 2D plots (histograms, networks, scatter plots...) coupled to 3D graphics for data analysis.
• Web services combining online database search and computations with desktop analysis.

Related Core TR&D projects

• TR&D 1: ChimeraX Multiscale Modeling Environment
• TR&D 2: Advanced Software for Modeling Molecular and Cellular Structures

Nuclear pore electron microscopy

• Nuclear pore composed of 30 different types, 1000 copies, 8-fold symmetry.
• GPU programming allows casting ambient lighting from all directions for better depth perception.

Human nuclear pore EM, 23 Angstroms resolution (Beck lab). Ambient occlusion lighting shadows crevices.	Simple lighting used by current generation visualization programs, hard to perceive depth.	Ribosome size compared to nuclear pore.

Nuclear pore atomic model 736 copies of 20 proteins, 4 million atoms composing about half the mass of nuclear pore fit to electron microscopy (Hoelz lab, 2016). Hierarchical organization: inner and outer rings, spokes, and Y-complex. Asymmetric unit has 92 proteins. Copies of Y-complex form scaffold for outer rings. Look in detail at integrative model of Y-complex.

Integrative hybrid model of Y-complex 7 protein Y-complex (also called nup84) forms scaffold for nuclear pore outer rings. Model built from many data sources: 2 protein X-ray models 5 comparative models, 16 template structures and sequence alignments 300 chemical crosslinks resolved by mass-spectroscopy, 2 different linkers 2D electron microscopy constraints model projection Disordered regions modeled with beads representing 10 residue segements Results Coarse grain models, one or more residues per bead. Ensemble of several thousand models, grouped in two clusters. Localization density maps summarize variation in protein localization. Data is saved in an mmCIF Integrated Hybrid Model format being developed in collaboration with RCSB for archiving at PDB.

Sequence alignments and templates for comparative models

• Comparative model quality can be judged from source sequence alignments and templates.
• Templates can be automatically fetched and aligned based on sequence alignments.

Models from IHM file	Template sequence alignments for one protein.	Superimposed template models on comparative model.

Model ensemble from sampling and scoring restraints Integrated Modeling Platform (IMP) software (Sali lab) computes ensembles of thousands of models scored based on restraints. Two clusters for two conformations each of contain 1000 models. Playing ensemble like dynamics trajectory shows variation within ensemble. Localization density maps for each protein in ensemble also shows variability.

Crosslink analysis

• Look at crosslink restraint satisfaction using interactive 2D plots.
• Crosslink length plot, mouse over histogram highlights 3D crosslinks.
• Crosslink ensemble length, mouse over histogram shows associated ensemble structure.

Crosslink network shows most crosslinks are intra-protein.

Lengths of 286 crosslinks for best model Length of one long crosslink for 1259 ensemble models. Is violated crosslink satisfied in any ensemble model?

Pull in additional protein-protein interface data using BLAST web service

• Is the protein interface correct for proteins spanned by very long crosslink?
• Use BLAST sequence search web service to find an x-ray structure with domains spanning the protein interface.
• BLAST nup133 (distal protein) sequence against PDB, low cutoff to find distant matches.
• Use PDB 3I4R as a bridge to align proteins nup84 to nup133.

BLAST PDB nup133 sequence	3ikr aligned to nup84 (aka nup107)	3ikr (light blue) bridging nup84/nup133

Summary

• ChimeraX allows visualization and analysis of diverse data types used in modeling large molecular machines:
  • X-ray / NMR structures
  • Comparative models
  • Sequence alignments
  • Chemical crosslinks
  • 3D and 2D electron microscopy
  • Coarse grain models, disordered segments
  • Ensembles of structures
• Advanced graphics, user interfaces including interactive 2d plots, web services, and unique analysis tools (e.g. secondary-structure guided alignment) enable visualization and analysis of complex molecular structures.