QSAR Prediction Model to Search for Compounds with Selective Cytotoxicity Against Oral Cell Cancer

Area of Science:

• Computational chemistry
• Drug discovery
• Cheminformatics

Background:

• Anticancer drugs often exhibit significant toxicity to both tumor and normal cells.
• Natural products can display high tumor specificity, offering potential therapeutic advantages.
• Previous research has identified cytotoxic activity and tumor specificity in various chemical compounds.

Purpose of the Study:

• To construct a database of previously reported compound data.
• To develop predictive models for screening novel anticancer drugs.
• To analyze structure-activity relationships for improved drug design.

Main Methods:

• Collected and merged data from 494 compounds, including activity and chemical structure.
• Utilized the random forest method to build predictive models for cytotoxicity and tumor specificity.
• Validated model performance using an external dataset.

Main Results:

• The structure-toxicity relationship prediction model demonstrated higher accuracy compared to the tumor selectivity model.
• Key molecular descriptors contributing to toxicity varied between tumor and normal cells.
• Established a comprehensive database for anticancer compound analysis.

Conclusions:

• Further research is necessary to enhance the predictive accuracy of tumor-selective toxicity models.
• Improved predictive models are anticipated to aid in the screening of candidate anticancer drugs.
• This work contributes to the ongoing effort to discover more targeted and effective cancer therapies.