FRnet-DTI: Deep convolutional neural network for drug-target interaction prediction

Abstract

The task of drug-target interaction prediction holds significant importance in pharmacology and therapeutic drug design. In this paper, we present FRnet-DTI, an auto encoder and a convolutional classifier for feature manipulation and drug target interaction prediction. Two convolutional neural neworks are proposed where one model is used for feature manipulation and the other one for classification. Using the first method FRnet-1, we generate 4096 features for each of the instances in each of the datasets and use the second method, FRnet-2, to identify interaction probability employing those features. We have tested our method on four gold standard datasets exhaustively used by other researchers. Experimental results shows that our method significantly improves over the state-of-the-art method on three of the four drug-target interaction gold standard datasets on both area under curve for Receiver Operating Characteristic(auROC) and area under Precision Recall curve(auPR) metric. We also introduce twenty new potential drug-target pairs for interaction based on high prediction scores.

Predicting Drug-Target Interactions with Electrotopological State Fingerprints and Amphiphilic Pseudo Amino Acid Composition

“…Docking simulation is the most successful method in drug-target interaction prediction when a three dimensional native structure of the target protein is available [ 10 ]. However, it is a time-consuming and expensive process to determine the native structure of a protein by sophisticated methods like X-ray Crystallography [ 11 ]. Thus, the 3D-structure of proteins are often unavailable.…”

“…These subdatasets have been used in a large number of related papers [ 14 , 15 , 16 , 17 , 18 , 19 ]. Due to the small number of known validated interactions among drug-target pairs, the unlabeled interaction pairs are considered as negative samples in most research and thus they outnumber positive samples [ 6 , 11 , 20 , 21 , 22 ]. The imbalance of DTI datasets is a major problem in supervised learning [ 11 , 23 ].…”

“…Due to the small number of known validated interactions among drug-target pairs, the unlabeled interaction pairs are considered as negative samples in most research and thus they outnumber positive samples [ 6 , 11 , 20 , 21 , 22 ]. The imbalance of DTI datasets is a major problem in supervised learning [ 11 , 23 ]. Another challenge is the representation of the drug-target pair which is often generated from molecular fingerprints of drugs and sequence or structure-based information of proteins [ 11 ].…”

“…The imbalance of DTI datasets is a major problem in supervised learning [ 11 , 23 ]. Another challenge is the representation of the drug-target pair which is often generated from molecular fingerprints of drugs and sequence or structure-based information of proteins [ 11 ].…”

DNN-DTIs: improved drug-target interactions prediction using XGBoost feature selection and deep neural network

“…A convolutional neural network (CNN)-based method proposed by Öztürk et al [30], using only sequence information and performing DTIs prediction on Davis and KIBA dataset with the AUPR values of 0.714 and 0.788, respectively. Rayhan et al [31] proposed the FRnet-DTI, using autoenconder and CNN for feature extraction and classification. The AUC and AUPR values on the gold standard dataset showed that FRnet-DTI could significantly boost the DTIs prediction performance.…”