Related Experiment Video
Updated: Jan 10, 2026
08:54
Facile Preparation and Photoactivation of Prodrug-Dye Nanoassemblies
Published on: February 17, 2023
1.5K
Machine Learning-Assisted Design and Discovery of High-Performance Cyanine-Based Photosensitizers for Integrated
Bowen Diao1, Shaoyang Shi1, Junhan Li1
1State Key Laboratory of Fine Chemicals, Frontiers Science Center for Smart Materials, Dalian University of Technology, Dalian, 116024, China.
Advanced Materials (Deerfield Beach, Fla.)
|November 20, 2025
Summary
Machine learning accelerates the design of cyanine photosensitizers for photodynamic therapy (PDT). This approach accurately predicts key properties, enabling efficient screening and identification of potent compounds for improved therapeutic outcomes.
Area of Science:
- Medicinal Chemistry
- Computational Chemistry
- Photodynamic Therapy
Background:
- Cyanine-based photosensitizers offer excellent near-infrared absorption and fluorescence for photodynamic therapy (PDT).
- Their clinical application is limited by low singlet oxygen quantum yield (ΦΔ) and challenges in structural optimization.
Purpose of the Study:
- To develop a machine learning (ML)-assisted molecular design framework for cyanine derivatives.
- To create predictive models for singlet oxygen quantum yield (ΦΔ) and fluorescence quantum yield (ΦF).
- To efficiently screen a large library of cyanine candidates for improved PDT agents.
Main Methods:
- Integrated RDKit structural and quantum chemical descriptors to build hybrid feature-based predictive models.
- Developed a two-stage virtual screening strategy to identify promising cyanine derivatives.
- Synthesized and validated three lead compounds, including a high-performing lead compound (1775).
Main Results:
- Achieved high prediction accuracy for ΦΔ and ΦF with R² > 0.9.
- Successfully screened 2835 candidate structures to identify potent cyanine derivatives.
- The lead compound 1775 demonstrated a high ΦΔ of 0.62 and performed well in cellular assays.
Conclusions:
- The ML-assisted framework provides a reliable and generalizable approach for rational design and rapid evaluation of cyanine-based theranostic agents.
- This data-driven paradigm bridges molecular modeling and experimental verification, accelerating the development of novel PDT drugs.
- The study validates the practical utility and robustness of ML in guiding experimental validation for drug discovery.
