Network-based cause–effect relationships. (a). Predicting the response of the human body to medication requires an understanding of drug–effect relationships at the organism, organ, tissue, cellular and molecular level, each having vastly different scales and structural complexities. Obtaining these drug-effect relationships requires comparison of large amounts of heterogeneous drug-effect information, which poses a formidable analytical challenge. (b) Current drug-discovery platforms are target-based and rely on observations of clinical effects to link drugs (i) and targets (ii) with organism effects (iv). This review examines protein-protein interaction network models (iii) and their use in medicine cause–effect analyses.
Components of bimolecular networks. (a) Proteins interact in a cell with other proteins, DNA, RNA and other molecules to form ‘biological molecular machines’. Their interactions include transient and stable complexes as well as physical and functional interactions. The protein interaction network that makes up an organism is probably more complex than the genome that underlies it. (b) Phenotype networks. Protein–protein interaction models have been created to identify putative networks associated with disease effects and drug response. By linking disease  and drug-induced  protein network topologies to effect network topologies, protein-centered, cause–effect models can be generated.
Link for Figure: http://rinalyzer.de/docu/rindata_gen.php
Introduction<br />Current platforms in understanding relationships between molecular structure and biological effects are structure-centered<br />Dogma- selectivity imparts efficacy and safety is not supported<br />The reality: target-based drug-discovery platforms are not able to predict drug-efficacy and the full spectrum of drugs in organisms<br />Complexity after molecular interaction is extra-ordinarily complex<br />Drug action: a coordinated response to multiple perturbations of cellular networks<br />Thus, there is a need of shift from structure/target-based platform to system/network-based platform <br />Challenging?<br />Fundamental rethinking of tranditional structure-function model<br />Biological effects are not defined by intended effects of a medicine or primary effects of disease<br />
Information flow in analysis of cause-effect relationships<br />Information on heterogenous molecular interactions exhibit complex regulatory scheme<br />Unlike reductionist approach, system-based strategies involve information flow within cellular and organism network systems<br />System-based cause-effect analysis employ network topology models<br />Disruption and production of macromolecular interactions: root of many diseases<br />
Overview of biological networks and network targets<br />Nodes (discrete molecules) and edges (functional connections)<br />Hubs – Nodes with higher no. of functional connections<br />Network properties – Scale free, robust, correlated (between network distance and functional distance)<br />Disease targets – bridging nodes<br />Medicine effects<br />
General steps in Network-based cause-effect analysis ( Construction of PPI network models)<br />Identification of network-reachable proteins<br />Network reachability – term used to identify proteins involved with discrete event<br />Strategy : identify known disease genes and map to proteins (on average a disease is associated with 12 genes)<br />Alternatively by the use of Analytical techniques (Quantitative MS, Phospho-peptide enrichment, aptamer technology, micro-western arrays, literature text mining) <br />Identification of cellular components capable of interacting with network-reachable proteins<br />Curated protein interaction databases<br />Problem: data variability<br />Solution: Integration of orthogonal molecular information and computational models based on genomic and structural information<br />Assessing viable routes for information flow between network-reachable proteins<br />Protein associations can be obtained by sorting protein-reachability profiles of drugs<br />Functional coupling<br />Through curated protein databases<br />
PPI disease networks<br />Most diseases are multifactorial<br />Use of PPI models for considering cause-effect associations is complicated<br />PPI networks models are refined whenever possible<br />Examples<br />Huntington’s disease (Huntington gene + GPTase)<br />Ataxia<br />Properties of disease networks<br />Diseased genes are not randomly positioned in a network<br />Disease states – resistant to perturbations<br />Protein hubs – highly conserved<br />Many diseases are results of small defects in many genes than large defects in few<br />
An example: Etiology of ataxia (lim et al., 2006)<br />PPI ataxia sub-network<br />
Uses for disease network topology<br />Identifying drug targets<br />Betweenness centrality and degree centrality - pharmacologically more significant<br />Less connected nodes affecting pathways -> more attractive drug targets<br />Targeting bridging nodes to identify potential drug targets<br />Multiple network targets<br />Attractive strategy for modifying phenotypes<br />Eg: non steroidal antiinflammatory drugs, antidepressants, anticancer drugs<br />Challenge: link desired and undesired effects of such medicine<br />Drug combinations<br />To target multiple sites within the same protein or multiple nodes within a molecular network<br />Well known examples:<br />Three drug combinations of reverse-transcriptase and protease inhibitors to treat HIV-infection<br />Four drug combinations to treat non-Hodgkin’s lymphoma<br />Drug repurposing<br />Imatinib: originally for chronic myelogenous leukemia; also to other cancers<br />Finasteride: originally for treating enlargement of prostrate gland but also for treating male baldness<br />
Use of disease network topologies to understand pharmacology of medicines<br />Copyright 2011 discovery medicine<br />
Capturing information on drug effects<br />Computer readable side-effect resource (SIDER)<br />Text mining of mendilian database<br />Sorting medicine-effect profiles<br />
A PPI network view of medicine effects<br />Figure: Aligning protein interaction and drug-effect topologies of medicines<br />
Caevets and Outlook<br />Current models are unable to capture the dynamics of cellular signaling<br />Long-term strategies – personalised medicine<br />
Final Thoughts<br />Broad range of heterogenous protein interaction and drug-effect data in IFA provides a new avenue for investigating cause-effect relationships in drug discovery<br />Use of cause effect analysis in combination with predictive modeling experiments provides roadmap for identifying circuits regulating transitions between various protein network topologies involved in pharmacological outcomes and disease progressionUnable to capture the dynamics of cellular signaling<br />The more detail and accurate the network topoloties are, the more we understand our biology, i.e higher success rate of new medicines, easy targets<br />