Analyzing disease-associated protein structures with visual analytics.

Protein function is related to protein structure, and understanding disease-associated structural changes is critical both to understanding protein-mediated illness and to developing clinical therapies. We have developed a visual analytics tool called ContactWalker that uses molecular dynamics simulations to investigate the structural and physical differences between wild type and disease-associated protein structures. This tool has been used successfully to characterize the effects of the disease ataxia with vitamin E deficiency (AVED). We are now beginning to investigate mutations to other disease-associated proteins including the tumor suppressor protein p53.

INTRODUCTION AND BACKGROUND:

Understanding protein structure is crucial to molecular disease research; molecular dynamics (MD) simulation is an in silico technique that provides high-resolution time-series data about protein dynamic structure. Unfortunately, MD simulations produce orders of magnitude more data than X-ray crystallography or NMR; datasets of this size are, as a unit, often too large for conventional tools.

METHODS:

To address this problem, we developed an interactive-speed visual analytics tool called ContactWalker1 that uses multiple MD simulations to calculate structural and physical differences between wild type and disease-associated protein structures. ContactWalker provides interactive access to the MD data and displays differences in protein structure via graph layout and protein structure visualization.

RESULTS:

We initially used ContactWalker to investigate ataxia with vitamin E deficiency (AVED), a neurodegenerative disease caused by mutations to the α-tocopherol transfer protein1; by analyzing differences in interatomic contacts, we characterized the mutation-associated structural differences, rationalized those findings with experiment, and offered a molecular-level explanation for the effects of the disease. We have since added additional physical analyses to ContactWalker and are now be eginning to investigate other mutant protein systems, including the tumor-suppressor protein p53. Figure 1 shows ContactWalker analysis of the p53 Y220C mutant. Figure 1A shows networks of mutation-associated disrupted contacts running through the protein. The focus box highlights the most disturbed region of the protein. This region, when mapped onto the protein structure (Fig. 1B), aligns with the location of a ligand binding pocket found experimentally to stabilize and rescue the mutant protein and to restore function2.

Figure 1.

p53 residues stabilized (green) and destabilized (orange) by the Y220C mutation. (A) ContactWalker output. (B) Protein structure.

DISCUSSION:

The figure above represents a differential summary of tens of gigabytes and hundreds of nanoseconds of MD data. Through simple and powerful interaction techniques, ContactWalker is able to pinpoint and highlight a highly-effective experimentally-discovered p53 rescue site. ContactWalker’s differential graph method supported by streaming interactive access to MD simulations of almost any size is a novel approach to interactive protein analysis.3 Taking the current p53 efforts together with the previous investigation of AVED, ContactWalker’s visual analytic approach to disease investigation looks promising.