HormoNet: a deep learning approach for hormone-drug interaction prediction

Area of Science:

• Biochemistry
• Pharmacology
• Computational Biology

Background:

• Human endogenous hormones significantly influence drug efficacy through various interactions.
• Understanding hormone-drug interactions (HDI) is vital for advancing precision medicine and optimizing drug treatments.

Purpose of the Study:

• To develop a deep learning model, HormoNet, for predicting hormone-drug interaction pairs.
• To assess the risk level associated with predicted hormone-drug interactions.
• To provide a computational tool for understanding complex hormone-drug relationships.

Main Methods:

• Integrated features from hormone and drug target proteins using amino acid composition and pseudo amino acid composition.
• Utilized 30 physicochemical and conformational properties for protein representation.
• Applied synthetic minority over-sampling technique to address data imbalance.
• Constructed novel benchmark datasets for HDI prediction and risk assessment.

Main Results:

• HormoNet demonstrated high performance in predicting HDI pairs and their risk levels on newly constructed datasets.
• The model effectively integrates protein features for accurate interaction prediction.
• The study provides valuable insights into hormone-drug relationships.

Conclusions:

• HormoNet offers a promising deep learning approach for predicting hormone-drug interactions and their associated risks.
• The findings can aid in designing safer and more effective drug therapies.
• The developed datasets and source code facilitate further research in this area.