Predicting fluorescence to singlet oxygen generation quantum yield ratio for BODIPY dyes using QSPR and machine...

Area of Science:

• Photochemistry
• Materials Science
• Computational Chemistry

Background:

• Functional dyes are essential for theranostics, integrating fluorescence imaging and photodynamic therapy (PDT).
• Developing new theranostic dyes is challenging due to complex synthesis and screening processes.
• Heavy-atom-free boron dipyrromethene (BODIPY) compounds offer potential as functional dyes.

Purpose of the Study:

• To develop machine learning methods for predicting fluorescence and photosensitizing abilities of BODIPY compounds.
• To analyze the ratio of fluorescence quantum yield (Φ_Fl) to singlet oxygen quantum yield (Φ_Δ) in different solvent environments.
• To streamline the discovery of novel theranostic agents.

Main Methods:

• Quantitative Structure-Property Relationship (QSPR) models were built using over 5000 molecular descriptors.
• Machine learning algorithms including Multiple Linear Regression (MLR), Support Vector Regression (SVR), and Random Forest Regression (RFR) were employed.
• Models were trained and validated using data from over 70 BODIPY structures in polar and non-polar solvents.

Main Results:

• Robust QSPR models achieved high statistical performance (R² = 0.73-0.91) for predicting photochemical parameters.
• The models accurately predicted dye properties in both hydrophilic (acetonitrile) and hydrophobic (toluene) simulated environments.
• Key molecular descriptors influencing dye performance, such as Eig03_EA(dm), F01[C-N], and TDB06p, were identified.

Conclusions:

• Machine learning-based QSPR methods are effective for predicting critical photochemical properties of BODIPY photosensitizers.
• These computational approaches can significantly accelerate the design and development of advanced theranostic agents.
• The study highlights the potential of AI in optimizing the discovery pipeline for functional dyes.