BLUE RIBBON POSTER THERAPY: Rapid Prediction of Optical Absorption Property for Tetrapyrrole-Based Photodynamic and Photothermal Therapy Agents
Abstract
Purpose
Photodynamic therapy and photothermal therapy commonly rely on tetrapyrrole based photosensitizers, whose therapeutic efficacy is governed by their electronic absorption property in the visible and near infrared regions. Accurate prediction of absorption spectra for newly designed tetrapyrrole compounds is computationally expensive, which limits high throughput drug discovery. This study aims to develop a fast and reliable framework for predicting electronic absorption spectra of tetrapyrrole macrocycles to enable rapid screening of next generation phototherapy agents.
Methods
A curated dataset of 1437 tetrapyrrole molecules, including porphyrins, chlorins, and bacteriochlorins, was constructed with density functional theory calculated frontier orbital energies defined by the four-orbital model. Molecular structures were represented using Simplified Molecular Input Line Entry System derived molecular fingerprints. Automated machine learning pipelines were implemented to predict HOMO -1, HOMO, LUMO, and LUMO +1 energies. Predicted orbital energies were subsequently used to reconstruct full electronic absorption spectra across the ultraviolet to near infrared range.
Results
The optimized models achieved high predictive performance with coefficients of determination of 0.985 on validation data and 0.936 on independent test data. Mean absolute prediction errors were below 0.1 electron volts across diverse tetrapyrrole scaffolds and substituent classes. Reconstructed absorption spectra closely reproduced key spectral features, including Q band and B band wavelengths and relative intensities that are crucial to phototherapy activation. Model performance remained robust across electron donating, electron withdrawing, and p conjugated substituents.
Conclusion
This framework enables rapid and accurate prediction of tetrapyrrole absorption spectra without the need for time intensive excited state calculations. The method provides an effective computational screening tool for the rational design and optimization of novel photodynamic and photothermal therapy drug candidates.