Speaker
Descrizione
As part of the activities of THE WP2, we have focused our efforts on understanding the mechanisms underlying the FLASH effect, i.e., the sparing of healthy tissue resulting from Ultra High Dose Rate (UHDR) irradiation, possibly with pulsed radiation and/or spatial fractionation (micro-mini-beam), and on quantifying it for clinical translation. To this end, we have developed a strategy that combines data organization and analysis with multiscale modeling and implemented this strategy in a single platform, to be used as a reference for data collection and processing and for calculating dose modification factors that correct treatment planning for the FLASH effect.
Data collection was realized on a customized database for outcomes from experiments conducted at the ElectronFlash linac of Centro Pisano Flash RadioTherapy (CPFR) . The platform, based on XNAT technology and built with the support of the cascade call FLASH_IT and with the collaboration of all the experimental groups participating in Spoke1, maps diverse experimental workflows into a unified data scheme, where beam parameters and biological variables are linked together for easy search and selection. A significant effort was dedicated to standardizing heterogeneous data , addressing ambiguities in physical definitions and biological endpoints for both in vivo and in vitro experiments . Besides irradiation parameters (dose rate, spatial and temporal structure of the beam, …), the platform includes structural and environmental variables of tissues and cells (strain, age, type of tissue, cell line/cycle, physiological/pathological state…), experimental outputs (clonogenic assays, toxicity scores, etc.), and files of diverse structure and formats with additional information (e.g., microscopy or MRI images, Excel or text files).
The integration of these data with modeling occurs in two directions: data are used to optimize models, while simulations deliver additional data to fill the gaps of experiments . More specifically, we performed extensive atomistic Monte Carlo, and Reactive Molecular dynamics , to assess the role of oxygen in the early phases of radical creation and to simulate the damage on DNA , on time and size scales difficult to access experimentally. These data are used together with experimental ones for the parameterization of low-resolution models with a higher empirical level, to simulate the evolution of damage on larger space and time scales , for more direct comparison with experiments. The combination of experimental and simulation data shed light on the mechanisms of FLASH effect, which, aside non-linear tracks interaction effects enhanced by oxygen, call into play a main role of differential diffusion and evolution of several agents, not only ROS and radicals but also scavengers, metabolic messengers and other molecules, which are currently under study, to clarify further mechanistic details.
In parallel, we developed a minimalistic empirical model specifically aimed at distilling the information (both experimental and simulation) and targeting directly the final effect, possibly with the aid of machine learning based algorithms . The model is built with the following ingredients: (i) minimal number of parameters (ii) two stages describing the local (sub-cellular) and the cell level damage (possibly extendable to tissue level) (iii) inclusion of stochastic-thresholds mechanisms and diffusivity-evolution of damaging agents. Our model reproduces the FLASH sparing effect as the dose rate is increased and incorporates the time dynamics and spatial fractionation of the beam. The simulations performed with the model indicate that the FLASH effect appears directly dependent on diffusivity, an intrinsic property of the cell/tissue type/status, while the temporal dynamics of delivery and space fractionation (beam characteristics) seem to modulate it. A main advantage of this model is the low number of parameters, which allows its optimization using a small number of data and can therefore be used, once optimized for different tissues/cell types, to evaluate the dose modifying factor to correct treatment planning systems for FLASH effect.
In conclusions, we designed and implemented a platform for data collection on FLASH effect from various origins (experimental and simulation) and in a variety of formats, expandable and customizable, which allow searching sorting and selecting data for further analysis. The platform is being populated with data and will remain beyond the THE project, and will be made available to the community. We performed simulations studies revealing, aside a possible amplification of non linear local irradiaton effect due to oxygen, a main role of diffusivity of damaging agents in the FLASH effect. These aspects are currently under analysis with further simulations. In the meanwhile, we addressed the calculation of the dose modifying factor with a minimalistic empirical stochastic model that includes the main determinants (mainly diffusivity / evolution of damaging agents) and is able to include the beam time pulsation and space fractionation. Once parameterized for different tissues, this model can efficiently and directly evaluate the dose modifying factor for the treatment planning.
