Identifying dysregulated pathways in cancers from pathway interaction networks

Abstract

Background: Cancers, a group of multifactorial complex diseases, are generally caused by mutation of multiple genes or dysregulation of pathways. Identifying biomarkers that can characterize cancers would help to understand and diagnose cancers. Traditional computational methods that detect genes differentially expressed between cancer and normal samples fail to work due to small sample size and independent assumption among genes. On the other hand, genes work in concert to perform their functions. Therefore, it is expected that dysregulated pathways will serve as better biomarkers compared with single genes.

Results: In this paper, we propose a novel approach to identify dysregulated pathways in cancer based on a pathway interaction network. Our contribution is three-fold. Firstly, we present a new method to construct pathway interaction network based on gene expression, protein-protein interactions and cellular pathways. Secondly, the identification of dysregulated pathways in cancer is treated as a feature selection problem, which is biologically reasonable and easy to interpret. Thirdly, the dysregulated pathways are identified as subnetworks from the pathway interaction networks, where the subnetworks characterize very well the functional dependency or crosstalk between pathways. The benchmarking results on several distinct cancer datasets demonstrate that our method can obtain more reliable and accurate results compared with existing state of the art methods. Further functional analysis and independent literature evidence also confirm that our identified potential pathogenic pathways are biologically reasonable, indicating the effectiveness of our method.

Conclusions: Dysregulated pathways can serve as better biomarkers compared with single genes. In this work, by utilizing pathway interaction networks and gene expression data, we propose a novel approach that effectively identifies dysregulated pathways, which can not only be used as biomarkers to diagnose cancers but also serve as potential drug targets in the future.

Figure 1

Schematic illustration of identifying dysregulated pathway in cancer. Firstly, gene expression profiles were standardized. Secondly, the genes were mapped to pathways. For each pathway, the principal component analysis (PCA) was employed to calculate the pathway activity score that summarizes the expression values of genes in each pathway. Thirdly, the pathway interaction network (PIN) was constructed based on gene expression data, protein-protein interactions, and cellular pathways. In the PIN, each node represents a pathway while each edge denotes the functional association between two pathways. Fourthly, the dysregulated pathways were identified as pathway markers that can best distinguish diseases from controls. The red node in PIN is the firstly identified pathway marker in disease, and the yellow ones are those pathway markers that can be combined with the first selected pathway to obtain best classification results while discriminating between diseases and controls.

Figure 2

Results obtained by PIN, PAC, BMI and gene biomarkers on four cancer datasets. Results obtained by PIN, PAC, BMI and gene biomarkers on four cancer datasets, where PIN, PAC, BMI and Gene respectively denotes our pathway biomarkers, PAC biomarkers, BMI biomarkers and gene biomarkers. (A). Lung cancer dataset, where PIN gets AUC score of 0.82 compared with 0.70 by PAC, 0.76 by BMI and 0.73 by Gene. (B). Prostate tumour dataset, where PIN gets AUC score of 0.82 compared with 0.71 by PAC, 0.77 by BMI and 0.63 by Gene. (C). Breast tumour dataset, where PIN gets AUC score of 0.99 compared with 0.92 by PAC, 0.93 by BMI and 0.90 by Gene. (D). Pancreatic tumour dataset, where PIN gets AUC score of 0.98 compared with 0.90 by PAC, 0.84 by BMI and 0.90 by Gene.

Figure 3

Results obtained by PIN, PAC, BMI and gene biomarkers on four lung cancer datasets. The biomarkers identified from lung cancer dataset (GSE 4115) by four methods were applied to independent lung cancer test datasets (GSE7670, GSE10072, GSE19027, and GSE2514), where PIN, PAC, BMI and Gene respectively denotes our pathway biomarkers, PAC biomarkers, BMI biomarkers and gene biomarkers. (A). GSE2514 dataset, where PIN gets AUC score of 0.99 compared with 0.99 by PAC, 0.95 by BMI and 0.87 by Gene. (B). GSE7670 dataset, where PIN gets AUC score of 0.99 compared with 0.99 by PAC, 0.80 by BMI and 0.85 by Gene. (C). GSE10072 dataset, where PIN gets AUC score of 0.99 compared with 0.99 by PAC, 0.93 by BMI and 0.96 by Gene. (D). GSE19027 dataset, where PIN gets AUC score of 0.71 compared with 0.63 by PAC, 0.65 by BMI and 0.52 by Gene.

Figure 4

Dysregulated pathways interaction network in pancreatic tumour dataset. In pancreatic tumour dataset (GSE16515), five dysregulated pathways were identified which can be assembled into a network based on their interactions in the pathway interaction network constructed for this dataset. Different colours were used to represent the five dysregulated pathways. The common genes between pathways are differentially expressed and the dashed line between two genes from distinct dysrugulated pathways denotes protein-protein interaction.