Multi-target drug repositioning by bipartite block-wise sparse multi-task learning

Abstract

Background: Finding potential drug targets is a crucial step in drug discovery and development. Recently, resources such as the Library of Integrated Network-Based Cellular Signatures (LINCS) L1000 database provide gene expression profiles induced by various chemical and genetic perturbations and thereby make it possible to analyze the relationship between compounds and gene targets at a genome-wide scale. Current approaches for comparing the expression profiles are based on pairwise connectivity mapping analysis. However, this method makes the simple assumption that the effect of a drug treatment is similar to knocking down its single target gene. Since many compounds can bind multiple targets, the pairwise mapping ignores the combined effects of multiple targets, and therefore fails to detect many potential targets of the compounds.

Results: We propose an algorithm to find sets of gene knock-downs that induce gene expression changes similar to a drug treatment. Assuming that the effects of gene knock-downs are additive, we propose a novel bipartite block-wise sparse multi-task learning model with super-graph structure (BBSS-MTL) for multi-target drug repositioning that overcomes the restrictive assumptions of connectivity mapping analysis.

Conclusions: The proposed method BBSS-MTL is more accurate for predicting potential drug targets than the simple pairwise connectivity mapping analysis on five datasets generated from different cancer cell lines.

Availability: The code can be obtained at http://gr.xjtu.edu.cn/web/liminli/codes .

Keywords: Drug repositioning; L1000; Multi-task learning.

Fig. 1

Illustration of the proposed model. Each block of A and B represents differential gene expression data after gene knock-down and drug treatment. W is the association matrix we would like to learn

Fig. 2

The calculated W of different methods for Data2.0 (1 ^st row), Data2.1 (2 ^nd row), Data2.2 (3 ^rd row) and Data2.3 (4^th row). Column (a): Ground truth W; (b): W connectivity mapping analysis test; (c): W ℓ₁ regularization; (d): ℓ_2,1 regularization; (e): BBSS-MTL with λ₃=0. BBSS-MTL performs best among all the methods in all simulated data set for learning the W

Fig. 3

The simulation results for BBSS-MTL with super-graph structure. (a): The structure similarity matrix K; (b, c, d): The perturbed Ws for Data2.1, Data2.2 and Data2.3; (e): The ground truth W; (f, g, h): The recovered W with K by BBSS-MTL. BBSS-MTL can recover the true W with the help of structure similarity, even when the datasets are perturbed

Fig. 4

The unified bipartite graph by BBSS-MTL across the three cell line datasets of HT29, MFC7 and PC3