Systems biology and machine learning approaches identify drug targets in diabetic nephropathy

Abstract

Diabetic nephropathy (DN), the leading cause of end-stage renal disease, has become a massive global health burden. Despite considerable efforts, the underlying mechanisms have not yet been comprehensively understood. In this study, a systematic approach was utilized to identify the microRNA signature in DN and to introduce novel drug targets (DTs) in DN. Using microarray profiling followed by qPCR confirmation, 13 and 6 differentially expressed (DE) microRNAs were identified in the kidney cortex and medulla, respectively. The microRNA-target interaction networks for each anatomical compartment were constructed and central nodes were identified. Moreover, enrichment analysis was performed to identify key signaling pathways. To develop a strategy for DT prediction, the human proteome was annotated with 65 biochemical characteristics and 23 network topology parameters. Furthermore, all proteins targeted by at least one FDA-approved drug were identified. Next, mGMDH-AFS, a high-performance machine learning algorithm capable of tolerating massive imbalanced size of the classes, was developed to classify DT and non-DT proteins. The sensitivity, specificity, accuracy, and precision of the proposed method were 90%, 86%, 88%, and 89%, respectively. Moreover, it significantly outperformed the state-of-the-art (P-value ≤ 0.05) and showed very good diagnostic accuracy and high agreement between predicted and observed class labels. The cortex and medulla networks were then analyzed with this validated machine to identify potential DTs. Among the high-rank DT candidates are Egfr, Prkce, clic5, Kit, and Agtr1a which is a current well-known target in DN. In conclusion, a combination of experimental and computational approaches was exploited to provide a holistic insight into the disorder for introducing novel therapeutic targets.

Figure 1

A schematic representation of the study design. This study was aimed at predicting novel drug targets for DN based on the holistic molecular pathogenesis map. Using different experimental and computational methods, the central nodes, key interactions, and signaling pathways of DN were identified. To translate the findings to clinical application, a high-performance machine learning framework, mGMDH-AFS, was developed and validated to predict drug targets for all human proteins. This classifier was then applied to candidate novel therapeutic targets in the constructed holistic map of DN. miRs: microRNAs; PPI: protein–protein interaction network.

Figure 2

The mouse model of DN was validated with different parameters. A mouse model of DN was established using streptozotocin and validated after 3 months using functional (a) and histopathological (b) assessments. Representative fields of normal (c) and DN (d, e) kidneys are shown. Data are reported as means ± SD. Asterisks represent P-value ≤ 0.05. GBM: glomerular basement membrane.

Figure 3

The kidney miRNA profile in DN. To assess microarray data quality in an unsupervised manner, principal component analysis and hierarchical clustering were performed (a). In microarray, miRNAs with |logFC|≥ 0.5 in cortex and medulla were determined (b). In addition to the miRNAs detected by microarray, some miRNAs which were experimentally shown or predicted to target DN-associated genes were selected (c). Among these candidates, 13 and 6 miRNAs were differentially expressed by qPCR in the cortex and medulla, respectively (d).

Figure 4

The holistic miRNA-target interaction maps were constructed, and key modules were identified. To investigate the role of differentially expressed miRNAs, the validated targets were identified, and the interaction networks were constructed with the differentially expressed miRNAs and their validated targets in the cortex (a) and medulla (b). Four modules in the cortex (I-IV) and one in medulla (V) networks were found, which potentially represent key interactions (c).

Figure 5

Key signaling pathways associated with DN. Pathway enrichment analysis with miRNA validated targets revealed 40 and 11 inter-connected pathways (adjusted P ≤ 0.05) for cortex (a) and medulla (b), respectively.

Figure 6

The novel machine could appropriately classify human proteins as drug targets or non-drug targets. The performance of the proposed mGMDH-AFS machine for drug target prediction based on biochemical or topology + biochemical features was acceptable and significantly superior to the examined standard machines including logistic regression (LG), radial basis function kernel support vector machine (RBF-SVM), generalized linear model (GLM), and radial basis function network (RBFN) as revealed by the hold-out validation. The proposed method significantly outperformed the state-of-the-art models (adjusted P ≤ 0.05). AUC: area under the receiver operating characteristic (ROC) curve; DOR: diagnosis odds ratio; MCC: Matthews correlation coefficient; DP: discriminant power.

Figure 7

The developed machine learning algorithm proposed novel therapeutic targets for DN. The performance of the mGMDH-AFS classifier to predict drug targets in the constructed networks of cortex and medulla was assessed by ROC curve analysis (a). The top predicted targets in cortex and medulla networks are shown either with biochemical or topology + biochemical features (b).