DFT-ML-Based Property Prediction of Transition Metal Complex Photosensitizers for Photodynamic Therapy

Photodynamic therapy (PDT) is a noninvasive clinical
treatment
for cancers using photosensitizers and light. While most research
has focused on organic molecules, such as porphyrins as photosensitizers,
there is emerging interest in the utilization of transition metal
complexes (TMCs). Photosensitizer synthesis and the following performance
test are time- and resource-consuming, so presynthetic screening of
photosensitizers for their property would be critical. In this work,
a hybrid mechanistic and data-driven model is proposed for the quantitative
structure–property relationship (QSPR) of photosensitizers;
important excited-state quantum chemistry descriptors (e.g., excitation
energy) are first calculated based on density functional theory (DFT),
and these descriptors, together with other molecular descriptors,
are used to build single and hybrid machine learning (ML) models for
the prediction of the singlet oxygen quantum yield of hexacoordinate
TMC photosensitizers (Ru-, Ir-, and Re-complex). The support vector
regression model and kernel ridge regression model are shown to provide
good predictions on test (R2 > 0.9)
and
external test sets (R2 > 0.7) in single-ML
models, while the delta-learning model and the Mixture-of-Experts
model can further improve the generalization ability (R2 up to 0.87 on the external test set) and show strong
universality. SHAP analysis further confirms the reasonable choice
of the mechanistic descriptors in the QSPR model. To our knowledge,
this constitutes the first integrated DFT-ML framework specifically
designed for the unique challenges of small data sets in TMC photosensitizer
research.