drugCIPHER

This repository contains the implementation of drugCIPHER, a linear regression framework to predict drug-target relations. For more details please see http://journals.plos.org/plosone/article?id=10.1371/journal.pone.0011764.

Introduction

drugCIPHER is a linear regression framework to integrate heterogenous drug similarities with protein interation network data to acurately predict drug-target relations. Three linear regression models are proposed respectively using drug therapeutic similarity, chemical similarity and their combination as responses and network distance as predictors.

Code

The code folder contains four matlab scripts for drugCIPHER.

• drugCIPHER_SingleS_Validation.m performs leave-one-out cross-validation on drug target prediction with single drug similarity matrix as input
• drugCIPHER_SingleS_Overall.m performs drug target overall prediction with single similarity matrix for all input drugs. Candidate targets are treated as all genes/proteins in the protein interation network
• drugCIPHER_MS_Validation.m performs leave-one-out cross-validation with two similarity matrices
• drugCIPHER_MS_Overall.m performs drug target overall prediction with two drug similarity matrices