RadPhysBio: A Radiobiological Database for the Prediction of Cell Survival upon Exposure to Ionizing Radiation

Area of Science:

• Radiobiology
• Biophysics
• Computational Biology

Background:

• Existing radiobiological databases are insufficient for predicting cell survival after ionizing radiation exposure.
• Accurate prediction of cell survival requires understanding DNA damage mechanisms and radiobiological parameters.

Purpose of the Study:

• To develop a comprehensive radiobiological database, 'RadPhysBio', for predicting human cell survival.
• To analyze the relationship between radiation type, linear energy transfer (LET), DNA double-strand breaks (DSBs), and cell survival coefficients (α and β).
• To create a machine learning model for predicting radiobiological parameters.

Main Methods:

• Manual literature search of PubMed (1980-2024) for experimental ionizing radiation data.
• Utilized WebPlotDigitizer and Python for calculating linear quadratic (LQ) model coefficients (α, β) and initial DNA double-strand breaks (DSBs).
• Employed the Monte Carlo code MCDS to simulate complex DNA damage and fill data gaps.
• Developed a biophysical prediction model using machine learning.

Main Results:

• Calculated α/β values align with existing literature.
• Observed a correlation between α and LET, but not β.
• Found a positive correlation between DSBs and LET in experimental data.
• Machine learning model demonstrated good performance for predicting α, identifying LET as a key feature.

Conclusions:

• The 'RadPhysBio' database provides a valuable resource for predicting irradiated cell survival using LQ model coefficients.
• Machine learning and repair models enhance the database's applicability and user reach.
• LET is a critical factor influencing DNA damage and cell survival prediction.