DrugECs: An Ensemble System with Feature Subspaces for Accurate Drug-Target Interaction Prediction

Area of Science:

• Computational drug discovery
• Bioinformatics
• Cheminformatics

Background:

• Drug-target interaction is crucial for identifying novel lead compounds.
• Traditional methods for finding lead compounds are time-consuming and error-prone.
• Computational techniques significantly reduce time and cost in drug design.

Purpose of the Study:

• To develop an efficient computational system for predicting drug-target interactions.
• To improve the accuracy and speed of lead compound identification.

Main Methods:

• Drug-target pairs encoded using a fragment technique for proteins and PaDEL-Descriptor software for drugs.
• Protein sequences divided into ordered fragments, encoded by amino acid physiochemical properties.
• Ensemble system built using k-Nearest Neighbor (kNN) classifier on resampled datasets.

Main Results:

• The proposed system effectively encodes drug and target molecules.
• Ensemble kNN classifier trained on overlapping subsets of drug-target pairs.
• Experimental validation on a drug-target dataset.

Conclusions:

• The developed prediction system outperforms current state-of-the-art methods.
• The system offers a faster and more accurate approach to predicting drug-target interactions.
• This advancement aids in accelerating the drug discovery pipeline.