Atomistic Visualization of Mesoscopic Whole-Cell Simulations

Abstract

Molecular visualizations are a principal tool for analyzing the results of biochemical simulations. With modern GPU ray casting approaches it is only possible to render several millions of atoms at interactive frame rates unless advanced acceleration methods are employed. But even simplified cell models of whole-cell simulations consist of at least several billion atoms. However, many instances of only a few different proteins occur in the intracellular environment, which is beneficial in order to fit the data into the graphics memory. One model is stored for each protein species and rendered once per instance. The proposed method exploits recent algorithmic advances for particle rendering and the repetitive nature of intracellular proteins to visualize dynamic results from mesoscopic simulations of cellular transport processes. We present two out-of-core optimizations for the interactive visualization of data sets composed of billions of atoms as well as details on the data preparation and the employed rendering techniques. Furthermore, we apply advanced shading methods to improve the image quality including methods to enhance depth and shape perception besides non-photorealistic rendering methods.