Multi-label classification using deep belief networks for virtual screening of multi-target drug

Abstract

Nowadays, the trend of drugs leads to multi-target drug. A drug compound may have one or more protein targets. Drugs that have multi-target protein considered to be more potential in the future. Virtual screening (VS) is a computational technique used in drug discovery to find the protein target of drugs. Virtual screening is usually based on compound similarity or database docking. Thus, the identification for multi-target drug compounds based on structure classification still remain as a challenging task The identification problem of multi-target protein from drug compounds can be categorized into multi-label classification problem. The purpose of this research is to find a new approach for multi-target drug virtual screening using machine learning technique. In this paper, the classification has been done by using combination of Deep Belief Networks (DBN) and Binary Relevance data transformation method. This research used two subset of protein target classes from DUD-E docking website. Feature were obtained from molecular fingerprint descriptor. The experiments result show that DBN can be used as virtual screening method for multi-target drug and outperform the DUD-E benchmarking.