Prediction of bioactive compound pathways using chemical interaction and structural information

Area of Science:

• Computational chemistry
• Biomedicine
• Pharmacology

Background:

• Functional screening of compounds is crucial in chemistry and biomedicine.
• Identifying compound pathways informs their correct application and biological effects.

Purpose of the Study:

• To develop and evaluate a computational method for identifying bioactive compound pathways.
• To integrate both structural and interaction information for improved pathway prediction.

Main Methods:

• A novel computational method was proposed, utilizing compound structural and interaction data.
• The method was evaluated using a dataset of 1,832 bioactive compounds from Selleckchem.
• Jackknife analysis was employed to assess the prediction accuracy.

Main Results:

• The proposed method achieved a total accuracy of 61.79%.
• This performance significantly surpassed other machine learning algorithms, which had accuracies below 46%.
• Analysis of false positives explored potential new pathway annotations for compounds.

Conclusions:

• The integrated computational approach effectively predicts bioactive compound pathways.
• The method offers a promising advancement over existing techniques relying solely on structural information.
• Further investigation into false positives may reveal novel compound-pathway associations.