Predicting DNA damage yields and assessing beam quality for protons and carbon ions using a DBSCAN algorithm.

Modeling radiation-induced DNA damage is essential for understanding the relative biological effectiveness (RBE) of ionizing radiation. This study presents a simplified framework using the density-based spatial clustering of applications with noise (DBSCAN) algorithm and using simulated physical track structures to characterize DNA damage induced by 0.5-200 MeV protons. The model assumed that (i) energy deposition ≥ 17.5 eV induced DNA damage, (ii) at least two damage points within a distance ε formed a cluster, (iii) isolated damage points were treated as noises, and (iv) the cluster-to-noise ratio corresponded to the double strand break (DSB)-to-single strand break (SSB) yield ratio. From the clustering output, a new beam quality metric called Quality of Beam (QoB; clusters per particle per µm) and its normalized form (clusters per keV of deposited energy) were defined. For protons, normalized QoB exhibited a strong linear correlation with DSB yields, enabling direct estimation of DSB and SSB yields using a single conversion factor. Applying the same framework and model parameters to carbon ions revealed a similar linear relationship between normalized QoB and DSB yields up to LET values of 200 keV µm⁻¹, beyond which the overkill effect emerged. The normalized QoB qualitatively reproduced RBE-LET trends and offered a biologically meaningful alternative to conventional metrics such as LET. Compared to full water radiolysis modeling, the DBSCAN framework was computationally efficient and provided a robust method for assessing ion beam quality and predicting DNA damage yields.