Dual-Module Near-Infrared Fluorophores Discovery System via Knowledge Transfer

Area of Science:

• Biomedical research
• Optical imaging
• Computational chemistry

Background:

• Near-infrared (NIR) imaging is crucial for in vivo studies due to deep tissue penetration and low autofluorescence.
• Discovering novel NIR fluorophores is challenging and labor-intensive.

Purpose of the Study:

• To develop a deep learning system for suggesting potential NIR fluorophores.
• To overcome limitations of traditional experimental methods in fluorophore discovery.

Main Methods:

• A dual-module deep learning framework was employed, featuring a predictive module with transfer learning and a generative module.
• The system estimates key optical properties and designs synthetically accessible NIR fluorophore candidates.
• Transfer learning was utilized to address data scarcity issues.

Main Results:

• Three novel NIR fluorophore candidates (NTDT-TPA, NPA-BTD, DPP-TPA) were synthesized and validated.
• NTDT-TPA, formulated into nanoparticles, demonstrated potential for in vitro and in vivo fluorescence bioimaging.
• The deep learning system successfully predicted optical properties and generated viable fluorophore structures.

Conclusions:

• The developed deep learning system accelerates the discovery of NIR fluorophores.
• NTDT-TPA shows significant potential as a fluorescent probe for in vivo bioimaging.
• This computational approach offers a more efficient alternative to traditional experimental screening for novel fluorophores.