Identifying High-Quality Leads among Screened Anticancerous Compounds Using SMILES Representations.

Area of Science:

• Computational chemistry and cheminformatics
• Artificial intelligence in drug discovery
• Oncology and pharmaceutical research

Background:

• Cancer remains a significant global health challenge, with chemotherapy being a primary treatment.
• High failure rates in anticancer drug development are attributed to safety and efficacy concerns.
• Computer-aided drug discovery (CADD) offers a promising approach to identify drug leads with improved profiles.

Purpose of the Study:

• To develop and evaluate an artificial intelligence (AI) model for predicting anticancer activity.
• To leverage molecular representations like Simplified Molecular Input Line Entry System (SMILES) for property prediction.
• To identify key molecular scaffolds associated with anticancer properties.

Main Methods:

• Utilized a multilayer perceptron (MLP) model for classification of anticancer activity.
• Employed cancer growth inhibition data from NCI-60 and ChEMBL drug datasets.
• Extracted and analyzed frequent ring and scaffold patterns from SMILES representations.

Main Results:

• The developed MLP model achieved a classification accuracy of 0.92 in distinguishing anticancer from nonanticancer compounds.
• Identified the top 8 frequent scaffolds from the analyzed datasets.
• Benchmarking demonstrated superior performance of the MLP model compared to other machine learning algorithms.

Conclusions:

• AI-driven analysis of SMILES representations can effectively predict anticancer activity.
• The MLP model shows significant potential for accelerating the identification of novel anticancer drug leads.
• Understanding molecular patterns from SMILES is crucial for designing safer and more efficacious cancer therapeutics.