Visualization of solution sets from automated docking of molecular structures

Abstract

Aligning structures, often referred to as docking or registration, is frequently required in fields such as computer science, robotics and structural biology. The task of aligning the structures is usually automated, but due to noise and imprecision, the user often needs to evaluate the results before a final decision can be made. The solutions involved are of a multidimensional nature and normally densely populated. Therefore, some form of visualization is necessary, especially if users want to achieve higher level understanding, such as solution symmetry or clustering, from the data. We have developed a system that provides two views of the data. One view places focus on the orientation of the solutions and the other focuses on translations. Solutions within the views are crosslinked using various visual cues. Users are also able to apply various filters, intelligently reducing the solution set. We applied the visualization to data generated by the automated cryo-EM process of docking molecular structures into electron density maps. Current systems in this field only allow for visual representation of a single solution or a numerical list of the data. We evaluated the system through a multi-phase user study and found that the users were able to gain a better high-level understanding of the data, even in cases of relatively small solution sets.