Wanwei Zhang

Deciphering early development of complex diseases by progressive module network

There is no effective cure nowadays for many complex diseases, and thus it is crucial to detect and further treat diseases in earlier stages. Generally, the development and progression of complex diseases include three stages: normal stage, pre-disease stage, and disease stage. For diagnosis and treatment, it is necessary to reveal dynamical organizations of molecular modules during the early development of the disease from the pre-disease stage to the disease stage. Thus, we develop a new framework, i.e. we identify the modules presenting at the pre-disease stage (pre-disease module) based on dynamical network biomarkers (DNBs), detect the modules observed at the advanced stage (disease-responsive module) by cross-tissue gene expression analysis, and finally find the modules related to early development (progressive module) by progressive module network (PMN). As an application example, we used this new method to analyze the gene expression data for NOD mouse model of Type 1 diabetes mellitus (T1DM). After the comprehensive comparison with the previously reported milestone molecules, we found by PMN: (1) the critical transition point was identified and confirmed by the tissue-specific modules or DNBs relevant to the pre-disease stage, which is considered as an earlier event during disease development and progression; (2) several key tissues-common modules related to the disease stage were significantly enriched on known T1DM associated genes with the rewired association networks, which are marks of later events during T1DM development and progression; (3) the tissue-specific modules associated with early development revealed several common essential progressive genes, and a few of pathways representing the effect of environmental factors during the early T1DM development. Totally, we developed a new method to detect the critical stage and the key modules during the disease occurrence and progression, and show that the pre-disease modules can serve as warning signals for the pre-disease state (e.g. T1DM early diagnosis) whereas the progressive modules can be used as the therapy targets for the disease state (e.g. advanced T1DM), which were also validated by experimental data.

Diagnosing phenotypes of single-sample individuals by edge biomarkers

Network or edge biomarkers are a reliable form to characterize phenotypes or diseases. However, obtaining edges or correlations between molecules for an individual requires measurement of multiple samples of that individual, which are generally unavailable in clinical practice. Thus, it is strongly demanded to diagnose a disease by edge or network biomarkers in one-sample-for-one-individual context. Here, we developed a new computational framework, EdgeBiomarker, to integrate edge and node biomarkers to diagnose phenotype of each single test sample. By applying the method to datasets of lung and breast cancer, it reveals new marker genes/gene-pairs and related sub-networks for distinguishing earlier and advanced cancer stages. Our method shows advantages over traditional methods: (i) edge biomarkers extracted from non-differentially expressed genes achieve better cross-validation accuracy of diagnosis than molecule or node biomarkers from differentially expressed genes, suggesting that certain pathogenic information is only present at the level of network and under-estimated by traditional methods; (ii) edge biomarkers categorize patients into low/high survival rate in a more reliable manner; (iii) edge biomarkers are significantly enriched in relevant biological functions or pathways, implying that the association changes in a network, rather than expression changes in individual molecules, tend to be causally related to cancer development. The new framework of edge biomarkers paves the way for diagnosing diseases and analyzing their molecular mechanisms by edges or networks in one-sample-for-one-individual basis. This also provides a powerful tool for precision medicine or big-data medicine.

Edge biomarkers for classification and prediction of phenotypes

In general, a disease manifests not from malfunction of individual molecules but from failure of the relevant system or network, which can be considered as a set of interactions or edges among molecules. Thus, instead of individual molecules, networks or edges are stable forms to reliably characterize complex diseases. This paper reviews both traditional node biomarkers and edge biomarkers, which have been newly proposed. These biomarkers are classified in terms of their contained information. In particular, we show that edge and network biomarkers provide novel ways of stably and reliably diagnosing the disease state of a sample. First, we categorize the biomarkers based on the information used in the learning and prediction steps. We then briefly introduce conventional node biomarkers, or molecular biomarkers without network information, and their computational approaches. The main focus of this paper is edge and network biomarkers, which exploit network information to improve the accuracy of diagnosis and prognosis. Moreover, by extracting both network and dynamic information from the data, we can develop dynamical network and edge biomarkers. These biomarkers not only diagnose the immediate pre-disease state but also detect the critical molecules or networks by which the biological system progresses from the healthy to the disease state. The identified critical molecules can be used as drug targets, and the critical state indicates the critical point of disease control. The paper also discusses representative biomarker-based methods.

Gaussian graphical model for identifying significantly responsive regulatory networks from time course high-throughput data

With rapid accumulation of functional relationships between biological molecules, knowledge-based networks have been constructed and stocked in many databases. These networks provide curated and comprehensive information for functional linkages among genes and proteins, whereas their activities are highly related with specific phenotypes and conditions. To evaluate a knowledge-based network in a specific condition, the consistency between its structure and conditionally specific gene expression profiling data are an important criterion. In this study, the authors propose a Gaussian graphical model to evaluate the documented regulatory networks by the consistency between network architectures and time course gene expression profiles. They derive a dynamic Bayesian network model to evaluate gene regulatory networks in both simulated and true time course microarray data. The regulatory networks are evaluated by matching network structure with gene expression to achieve consistency measurement. To demonstrate the effectiveness of the authors method, they identify significant regulatory networks in response to the time course of circadian rhythm. The knowledge-based networks are screened and ranked by their structural consistencies with dynamic gene expression profiling.

MCentridFS: a tool for identifying module biomarkers for multi-phenotypes from high-throughput data

Systematically identifying biomarkers, in particular, network biomarkers, from high-throughput data is an important and challenging task, and many methods for two-class comparison have been developed to exploit information of high-throughput data. However, as the high-throughput data with multi-phenotypes are available, there is a great need to develop effective multi-classification models. In this study, we proposed a novel approach, called MCentridFS (Multi-class Centroid Feature Selection), to systematically identify responsive modules or network biomarkers for classifying multi-phenotypes from high-throughput data. MCentridFS formulated the multi-classification model by network modules as a binary integer linear programming problem, which can be solved efficiently and effectively in an accurate manner. The approach is evaluated with respect to two diseases, i.e., multi-stages HCV-induced dysplasia and hepatocellular carcinoma and multi-tissues breast cancer, both of which demonstrated the high classification rate and the cross-validation rate of the approach. The computational results of the five-fold cross-validation of the two data show that MCentridFS outperforms the state-of-the-art multi-classification methods. We further verified the effectiveness of MCentridFS to characterize the multi-phenotype processes using module biomarkers by two independent datasets. In addition, functional enrichment analysis revealed that the identified network modules are strongly related to the corresponding biological processes and pathways. All these results suggest that it can serve as a useful tool for module biomarker detection in multiple biological processes or multi-classification problems by exploring both big biological data and network information. The Matlab code for MCentridFS is freely available from http://www.sysbio.ac.cn/cb/chenlab/images/MCentridFS.rar.

Discovering a critical transition state from nonalcoholic hepatosteatosis to nonalcoholic steatohepatitis by lipidomics and dynamical network biomarkers

Nonalcoholic fatty liver disease (NAFLD) is a major risk factor for type 2 diabetes and metabolic syndrome. However, accurately differentiating nonalcoholic steatohepatitis (NASH) from hepatosteatosis remains a clinical challenge. We identified a critical transition stage (termed pre-NASH) during the progression from hepatosteatosis to NASH in a mouse model of high fat-induced NAFLD, using lipidomics and a mathematical model termed dynamic network biomarkers (DNB). Different from the conventional biomarker approach based on the abundance of molecular expressions, the DNB model exploits collective fluctuations and correlations of different metabolites at a network level. We found that the correlations between the blood and liver lipid species drastically decreased after the transition from steatosis to NASH, which may account for the current difficulty in differentiating NASH from steatosis based on blood lipids. Furthermore, most DNB members in the blood circulation, especially for triacylglycerol (TAG), are also identified in the liver during the disease progression, suggesting a potential clinical application of DNB to diagnose NASH based on blood lipids. We further identified metabolic pathways responsible for this transition. Our study suggests that the transition from steatosis to NASH is not smooth and the existence of pre-NASH may be partially responsible for the current clinical limitations to diagnose NASH. If validated in humans, our study will open a new avenue to reliably diagnose pre-NASH and achieve early intervention of NAFLD.

A Gaussian graphical model for identifying significantly responsive regulatory networks from time series gene expression data

With rapid accumulation of functional relationships between biological molecules, knowledge-based networks have been constructed and stocked in many databases. These networks provide the curated and comprehensive information for the functional linkages among genes and proteins, while their activities are highly related with specific phenotypes and conditions. To evaluate a knowledge-based network in a specific condition, measuring the consistency between its structure and the conditionally specific gene expression profiling data is an important criterion. In this work, we propose a Gaussian graphical model to evaluate the documented regulatory networks by the consistency between network architectures and time-series gene expression profiles. By developing a dynamical Bayesian network model, we derive a new method to evaluate gene regulatory networks in both simulated and true time series microarray data. The regulatory networks are evaluated by matching a network structure and gene expressions, which are achieved by randomly rewiring the regulatory structures. To demonstrate the effectiveness of our method, we identify the significant regulatory networks in response to the time series gene expression of circadian rhythm. Moreover, the knowledge-based networks are screened and ranked by their consistencies of structures based on dynamical gene expressions.