Systems Biology (Methods in Molecular Biology)

Analyzing the structure of biochemical disease networks pro-

vides useful information including network hubs, regulatory

motifs, and possibly global features like small world organization

of the system. Regulatory motifs, feedback, and feedforward loops

are a source for nonlinear regulatory behavior [4, 5], which not

only challenges human intuition but also limits the application of

conventional data analysis tools [6–8]. Moreover, for large-scale

biochemical networks a dynamical analysis is particularly difficult

with mechanistic (e.g., ODE-based) approaches from the theory of

dynamical systems. In order to exploit the advantages of large-scale

biochemical networks, in combination with mechanistic modeling,

we need integrative approaches and computational workflows to

identify disease-specific small regulatory/function modules that

can be subjected to a more detailed analysis, followed by the

prediction of molecular signatures. Exploring large-scale nonlinear

dynamical networks will remain an art form. What we are aiming for

here is a rational approach to what is effectively guesswork, forced

upon us by the wonderful complexity found in living systems.

In this chapter, we highlight and discuss an integrative work-

flow (Fig.1) to study large-scale biochemical disease networks by

combining techniques from bioinformatics and systems biology.

Integrating experimental and clinical data with the workflow,

process-specific hypotheses can be generated and validated. In

particular, we present here a flexible and extendible workflow that

combines network structural properties with high-throughput and

other biomedical data to identify smaller modules/molecular sig-

natures for tumor-specific disease phenotypes. A mathematical

model of the identified smaller modules/molecular signatures can

be constructed to give mechanistic understanding of the disease

and propose new hypotheses, which are subject to experimental

validation. For the illustration of the workflow, we used prostate

cancer as a case study with the aim of identifying key functional

components involved in regulation and progression of primary to

metastasis transitions [9]. We construct a network for each of the

clinical states of prostate cancer based on differentially expressed

Structural

analysis

Mathematical

modelling

constructionNetwork Validation

• Literature mining


• Databases


• Domain knowledge
  
  
  • Centrality measures
  
  
  • Network motifs
  
  
  • Gene prioritization
    
    
    • Network analysis
    
    
    • Biomedical properites
    
    
    • Gene expression data
    
    
    
    
  
  
  
  

Regulatory

network

elements

• Regulatory elements
  for prostrate cancer
  
  
  • ODE based model
  
  
  • Logic-based model
  
  
  • Hybrid model
  
  
  
  


• Predictions of disease
  molecular signatures


• Identification of potential
  drug targes


• Clinical data validation


• In vitro validation
  Multi-objective
  function

Graph theory

Interactions map - Validated molecular

signatures

• Validated drug targets

Fig. 1An integrative workflow to analyze large-scale biochemical disease networks

248 Faiz M. Khan et al.