Scalable prediction of compound-protein interactions using minwise hashing

Area of Science:

• Bioinformatics
• Cheminformatics
• Computational Biology

Background:

• Accurate prediction of compound-protein interactions is crucial for drug discovery and identifying therapeutic targets.
• Existing methods struggle with the scale of genome-wide interaction prediction.

Purpose of the Study:

• To develop a novel, scalable chemogenomic method for predicting compound-protein interactions.
• To improve prediction accuracy, computational efficiency, and model interpretability.

Main Methods:

• Utilized an improved minwise hashing algorithm to create compact fingerprints for compound-protein pairs.
• Applied a sparsity-induced classifier to these compact fingerprints for prediction.

Main Results:

• The proposed method demonstrated superior performance compared to previous chemogenomic techniques.
• Achieved higher prediction accuracy and computational efficiency on large datasets.
• The method is computationally feasible for millions of compound-protein pairs, unlike prior approaches.

Conclusions:

• The novel chemogenomic method offers a scalable and efficient solution for predicting compound-protein interactions.
• This approach is valuable for virtual screening of numerous compounds against diverse protein targets.
• Enables more effective drug lead discovery and identification of new therapeutic targets.