A new open-source GPU-based microscopic Monte Carlo simulation tool for the calculations of DNA damages caused by ionizing radiation --- Part I: Core algorithm and validation.

Purpose: Monte Carlo (MC) simulation of radiation interactions with water medium at physical, physicochemical, and chemical stages, as well as the computation of biologically relevant quantities such as DNA damages, are of critical importance for the understanding of microscopic basis of radiation effects. Due to the large problem size and many-body simulation problem in the chemical stage, existing CPU-based computational packages encounter the problem of low computational efficiency. This paper reports our development on a GPU-based microscopic Monte Carlo simulation tool gMicroMC using advanced GPU-acceleration techniques. Methods: gMicroMC simulated electron transport in the physical stage using an interaction-by-interaction scheme to calculate the initial events generating radicals in water. After the physicochemical stage, initial positions of all radicals were determined. Simulation of radicals' diffusion and reactions in the chemical stage was achieved using a step-by-step model using GPU-accelerated parallelization together with a GPU-enabled box-sorting algorithm to reduce the computations of searching for interaction pairs and therefore improve efficiency. A multi-scale DNA model of the whole lymphocyte cell nucleus containing similar to 6.2 Gbp of DNA was built. Results: Accuracy of physical stage simulation was demonstrated by computing stopping power and track length. The results agreed with published data and the data produced by GEANT4-DNA (version 10.3.3) simulations with 10 -20% difference in most cases. Difference of yield values of major radiolytic species from GEANT4-DNA results was within 10%. We computed DNA damages caused by monoenergetic 662 keV photons, approximately representing Cs-137 decay. Single-strand break (SSB) and double-strand break (DSB) yields were 196 +/- 8 SSB/Gy/Gbp and 7.3 +/- 0.7 DSB/Gy/Gbp, respectively, which agreed with the result of 188 SSB/Gy/Gbp and 8.4 DSB/Gy/Gbp computed by Hsiao et al. Compared to computation using a single CPU, gMicroMC achieved a speedup factor of similar to 540x using an NVidia TITAN Xp GPU card. Conclusions: The achieved accuracy and efficiency demonstrated that gMicroMC can facilitate research on microscopic radiation transport simulation and DNA damage calculation. gMicroMC is an open-source package available to the research community. (C) 2020 American Association of Physicists in Medicine.