Computer-aided prediction for medicinal chemistry via the Internet

Area of Science:

• Medicinal Chemistry
• Computational Chemistry
• Drug Discovery

Background:

• Numerous computational tools are available online for medicinal chemistry applications.
• Assessing the predictive accuracy of these freely available tools is crucial for reliable drug design.

Purpose of the Study:

• To compare the predictive performance of various freely accessible internet-based computational tools for medicinal chemistry.
• To evaluate the reliability of predictions for key physicochemical and biological properties.

Main Methods:

• Comparison of correlation coefficients for log P, aqueous solubility, and drug-likeness predictions across multiple computational methods.
• Evaluation of the PASS (Prediction of Activity Spectra) Inet tool for biological activity profiling.

Main Results:

• Correlation coefficients for log P ranged from 0.65 to 0.90 (seven methods), indicating reasonable accuracy.
• Correlation coefficients for aqueous solubility (0.01-0.73, four methods) and drug-likeness (0.19-0.73, three methods) were poor.
• PASS Inet demonstrated high accuracy in predicting biological activity, aligning well with experimental results for various compound series.

Conclusions:

• Medicinal chemists must validate internet-based computational tool predictions against experimental data, especially for aqueous solubility and drug-likeness.
• PASS Inet is a valuable tool for broad biological activity profiling in drug discovery.
• Computer-aided predictions offer advantages but also have limitations that require careful consideration.