BackgroundMinibeam radiation therapy (MBRT) is a novel technique which has been shown to widen the therapeutic window through significant normal tissue sparing. Despite the heterogeneous dose distributions, tumor control is still ensured. Nevertheless the exact radiobiological mechanisms responsible for MBRT efficacy are not fully understood.PurposeReactive oxygen species (ROS) resulting from water radiolysis were investigated given their implications not only on targeted DNA damage, but also for their role in the immune response and non‐targeted cell signalling effects: two potential drivers of MBRT efficacy.MethodsMonte Carlo simulations were performed using TOPAS‐nBio to carry out the irradiation of a water phantom with beams of protons (pMBRT), photons (xMBRT), 4He ions (HeMBRT), and 12C ions (CMBRT). Primary yields at the end of the chemical stage were calculated in spheres of 20 μm diameter, located in the peaks and valleys at various depths up to the Bragg peak. The chemical stage was limited to 1 ns to approximate biological scavenging, and the yield of · OH, H2O2, and was recorded.ResultsBeyond 10 mm, there were no substantial differences in the primary yields between peaks and valleys of the pMBRT and HeMBRT modalities. For xMBRT, there was a lower primary yield of the radical species · OH and at all depths in the valleys compared to the peaks, and a higher primary yield of H2O2. Compared to the peaks, the valleys of the CMBRT modality were subject to a higher · OH and yield, and lower H2O2 yield. This difference between peaks and valleys became more severe in depth. Near the Bragg peak, the increase in the primary yield of the valleys over the peaks was 6% and 4% for · OH and respectively, while there was a decrease in the yield of H2O2 by 16%. Given the similar ROS primary yields in the peaks and valleys of pMBRT and HeMBRT, the level of indirect DNA damage is expected to be directly proportional to the peak to valley dose ratio (PVDR). The difference in the primary yields implicates a lower level of indirect DNA damage in the valleys compared to the peaks than what would be suggested by the PVDR for xMBRT, and a higher level for CMBRT.ConclusionsThese results highlight the notion that depending on the particle chosen, one can expect different levels of ROS in the peaks and valley that goes beyond what would be expected by the macroscopic PVDR. The combination of MBRT with heavier ions is shown to be particularly interesting as the primary yield in the valleys progressively diverges from the level observed in the peaks as the LET increases. While differences in the reported · OH yields of this work implicated the indirect DNA damage, H2O2 yields particularly implicate non‐targeted cell signalling effects, and therefore this work provides a point of reference for future simulations in which the distribution of this species at more biologically relevant timescales could be investigated.