EVALUATION OF PHOTOSENSITIZERS FOR PHOTODYNAMIC THERAPY WITH MACHINE LEARNING TECHNIQUES
Good photosensitizers have high photochemical, electrochemical and thermal stability, which gives them diverse technological applications, such as dyes, in solar cells and in photodynamic therapy (PDT). PDT is used in the treatment of cancer and other non-oncological diseases, in which photosensitizing drugs activated by light at specific wavelengths are used. PDT destroys cancer cells, preserving healthy cells. However, the efficiency of PDT critically depends on the properties of the photosensitizer, driving the need to discover new compounds with ideal properties. The integration of quantum chemistry calculations and Machine Learning (ML) methods emerges as a powerful strategy for this discovery. This study proposed a new method for identifying photosensitizers using ML, performing quantum chemistry calculations to understand properties of the main dyes used in PDT. Furthermore, a database with detailed information on molecular properties was structured. Chemical descriptors were implemented and classification models were constructed, using CatBoost, LightGBM and XGBoost. The results showed robust and effective models in discriminating between dyes and non-dyes, indicating their applicability in screening dyes for photodynamic therapy.