Deep belief networks using hybrid fingerprint feature for virtual screening of drug design

Virtual screening (VS) is a computational technique used in drug discovery. VS process usually works by identifying the ability of structures to bind each other. One of the structure interpretation is molecular fingerprints. Molecular fingerprints are used for computational drug discovery as feature for VS. A variety of fingerprint types has been introduced. Combining two or more fingerprints into a hybrid fingerprints has been found to improve the performance of VS. Furthermore, machine learning techniques have helped to improve the performance of VS. The purpose of this research is to find a new Deep Belief Networks (DBN) architecture approach for hybrid fingerprint features. In this paper, there were two different approaches for combining two fingerprints feature for DBN, then called initial combining and latter combining. This research used six protein target classes as same as the previous research about DBN for VS. The experiments result show that the best way to combine the fingerprints for DBN architecture is initial combining.