How a cell decides whether to survive radiation express
A cell hit by a proton beam is like a city after a pinpoint bombing: some blocks are leveled, others just have broken windows. Whether the city survives depends on which buildings are hit and how the strikes are distributed. In the past, doctors gauged the damage by the overall intensity of the bombardment, but that's not enough: two identical doses can leave very different destruction.
The new model works like a meticulous damage assessor, looking not only at the force of the explosions but also at their locations. It analyzes how particle energy is distributed across different scales, from tiny clusters a few nanometers wide (where DNA strands break) to micrometer tracks (the trails left by particles in tissue). The model decides on its own which details to focus on and shows them to humans, like a translator who underlines key words in a document.
The most interesting part: no single scale turned out to be dominant. To make accurate predictions, you need all of them — from nanometer damage to micrometer tracks. But the smaller the scale, the greater its contribution. The usual LET metric (linear energy transfer — how much energy a particle deposits along each millimeter of its path) turned out to be secondary. It's like assessing bomb damage by which specific houses are hit rather than the total bomb yield.
This approach could change how radiation therapy is planned: instead of rough averages, doctors would get a map of a tumor's vulnerable spots. The model also helps explain why some cells die from radiation while others survive, and how to protect healthy tissue during treatment.
Original abstract
as written by the authors arXiv:2601.15106
The relationship between the physical characteristics of the radiation field and biological damage is central to both radiotherapy and radioprotection, yet the link between spatial scales of energy deposition and biological effects remains not entirely understood. To address this, we developed an interpretable deep learning model that predicts cell survival after proton and carbon ion irradiation, leveraging sequential attention to highlight relevant features and provide insight into the contribution of different energy deposition scales. Trained and tested on the PIDE dataset, our model incorporates, beside LET, nanodosimetric and microdosimetric quantities simulated with MC-Startrack and Open-TOPAS, enabling multi-scale characterization. While achieving high predictive accuracy, our approach also emphasizes transparency in decision-making. We demonstrate high accuracy in predicting RBE for in vitro experiments. Multiple scales are utilized concurrently, with no single spatial scale being predominant. Quantities defined at smaller spatial domains generally have a greater influence, whereas the LET plays a lesser role.
The energy deposited by a single particle in a cell can be concentrated in clusters just a few atoms across, and it's these clusters that most often break DNA.