Drug-target interaction prediction via an ensemble of weighted nearest neighbors with interaction recovery

Abstract

Predicting drug-target interactions (DTI) via reliable computational methods is an effective and efficient way to mitigate the enormous costs and time of the drug discovery process. Structure-based drug similarities and sequence-based target protein similarities are the commonly used information for DTI prediction. Among numerous computational methods, neighborhoodbased chemogenomic approaches that leverage drug and target similarities to perform predictions directly are simple but promising ones. However, most existing similarity-based methods follow the transductive setting. These methods cannot directly generalize to unseen data because they should be re-built to predict the interactions for new arriving drugs, targets, or drug-target pairs. Besides, many similarity-based methods, especially neighborhood-based ones, cannot handle directly all three types of interaction prediction involving new drugs and/or new targets. Furthermore, a large amount of missing (undetected) interactions in current DTI datasets hinders most DTI prediction methods. To address these issues, we propose a new method denoted as Weighted k-Nearest Neighbor with Interaction Recovery (WkNNIR). Not only can WkNNIR estimate interactions of any new drugs and/or new targets

Heterogeneous network propagation with forward similarity integration to enhance drug–target association prediction

“…This model combines DDA_Jac, DDI_Cos, and Structures as an integrated drug similarity using SNF and employs Seq_Loc as a target similarity. To verify that the model selected by FSI is the best, we compared its performance to those of the models with full similarity integration, random similarity integrations, and the original heterogeneous network (OHN) model, which uses only Structures and Seq_Loc, proposed by Liu et al (2022) , Liu et al (2016) , Peng et al (2022) and Wang, Yang & Li (2013a) . To demonstrate the efficiency of FSI, we first compare the performance of the FSI model with that of the full model which fuses all drug similarities and target similarities, as shown in Fig.…”

“…To predict DTIs, the network models utilize the drug–drug similarity derived only from drug chemical structures and the target–target similarity based only on local sequence alignments of target proteins. To show the superior performance of the model optimally integrating multiple drug and target similarities by FSI, we compared the performances of the OHN model ( Liu et al, 2022 ; Liu et al, 2016 ; Peng et al, 2022 ; Wang, Yang & Li, 2013a ) with those of the FSI model ( Fig. 7 ).…”

A Hybrid ROS-SVM Model for Detecting Target Multiple Drug Types

“…Liu Bin et al discuss drug target interaction using the Nearest Neighbor weighting technique and sampling the drug probabilities. However, in this study, it is not explained what the sampling method is only local sampling is mentioned [18]. Thafar Maha et al discuss drug prediction using a graph-shaped approach and name similarity.…”

Novel prediction methods for virtual drug screening

“…Also, most existing similarity-based methods need to be retrained each time new drugs are approved and they can't be directly used for novel drug discovery or when adding new targets. There are some novel methods like Weighted k-Nearest Neighbor with Interaction Recovery (WkNNIR) [43] that aim to solve those problems by using ensemble of DTI methods with different sampling strategies and interaction recovery to perform prediction upon a completed interaction matrix without the necessity of retraining and ability to detect previous false negatives.…”