ChemFLuo: a web-server for structure analysis and identification of fluorescent compounds

Area of Science:

• Chemical Biology
• Drug Discovery
• Computational Chemistry

Background:

• Fluorescent detection is crucial in chemical biology.
• Fluorescent compounds in chemical libraries can interfere with identifying genuinely active compounds.
• This interference is reproducible and concentration-dependent, necessitating reliable screening tools.

Purpose of the Study:

• To develop a webserver, ChemFLuo, for detecting fluorescent compounds.
• To provide a credible screening tool for early-stage drug discovery.
• To filter out compounds with undesirable fluorescent properties.

Main Methods:

• Developed prediction models using machine learning algorithms and molecular representations.
• Trained models on large datasets of blue and green fluorescent compounds.
• Identified representative fluorescent substructures for screening.

Main Results:

• The best blue fluorescence prediction model achieved a balanced accuracy (BA) of 0.858 and an area under the receiver operating characteristic curve (AUC) of 0.931 on the validation set.
• The best green fluorescence prediction model achieved a BA of 0.810 and an AUC of 0.887 on the validation set.
• Summarized 22 blue and 16 green fluorescent substructures.

Conclusions:

• ChemFLuo is a public webserver for filtering compounds with undesirable fluorescent properties.
• The tool aids in designing high-quality chemical libraries for drug discovery.
• The webserver is freely available at http://admet.scbdd.com/chemfluo/index/.