2013 Archived Content
SC4 Network Pharmacology
Part 1: Biological networks and complex systems
Biological networks have evolved to be very robust. This is a fundamental organizing principle of life. Biochemical and physiological networks need to be able to continue to function in the face of continual internal and external perturbation. In common with many other types of highly optimized, large-scale networks, they have the general property of attack tolerance. Random removal of individual components of a biological network has surprisingly little functional consequence for the network, even if it is in a pathological state. To have an effect, interventions within a complex biological network need to be multiple and highly selective. Biological networks that underlie systems-level functions (e.g. apoptosis, inflammation, autophagy) are structurally very complex, and so predicting the functional outcome of interventions or the consequence of mutations that must then be addressed pharmacologically in a disease is not trivial.
Networks are amenable to analysis using several branches of mathematics, including graph theory, Information theory and set theory. Biological networks can be represented as graphs (points [more commonly termed nodes or vertices] e.g. proteins, connected by lines representing interactions [called edges]); and the local and global properties of this topological map can be calculated. Similarly, community substructure and community node role can be inferred from information in the pattern of connections. This information can then be used to identify sets of high value nodes, some of which might serve as ‘targets’ for drugs, depending on the desired therapeutic outcome. We will discuss with examples the properties and analysis of networks of biological complexity with the aim of identifying “target sets” for new drugs.
Part 2: Chemical promiscuity
Drugs generally exert their effects through binding to proteins, thereby modulating their biophysical properties. Bioactive compounds invariably influence more than one protein, either as a consequence of structural similarities between the intended target and other proteins; or through allosteric effects on other proteins; or through pleiotropy, where an interaction results in multiple downstream effects on other proteins (e.g. through changes in phosphorylation state, methylation, post-transcriptional processing, expression and abundance of other proteins); or through multivalent target binding by different presentations of the active molecule. Furthermore, any active metabolite of the primary drug is subject to the same range of possible ways of interacting with many proteins in the body.
In whatever way this polyvalent interaction occurs, the net result is often unpredictable efficacy or safety, and often both. Occasionally, a drug’s promiscuity engages a fortuitous combination of appropriate high-value network targets in disease-related cells and simultaneously misses nodes of significance in normal cells, to produce a treatment success. The more highly specific and the less promiscuous a molecule is for a particular target the more important that target has to be in network terms for it to have a significant effect. We will discuss, using known examples, cases where the polyvalent footprint of drugs contributes either positively or negatively to their efficacy and safety properties.
Part 3: Network pharmacology:
Network pharmacology is the judicious application of the above principles to optimize the efficacy and safety of a drug molecule.
The first component of network pharmacology is the selection of optimal intervention points in the network(s) of relevance to a disease process. Having curated disease-related network data, combinatorial network impact algorithms can be applied to calculate the optimal combination of proteins to affect to modulate the network(s) maximally, while giving minimal or acceptable modulation of normal cell networks. This has been termed ‘combinatorial impact analysis’. “Impact” can be quantified by many different metrics but relates to how much the integrity of the network(s) can be changed by the interventions. Since disease-related networks typically relate directly to systems-level function or dysfunction, combinatorial impact can relate closely to impact on the disease. With this idealized “footprint” in mind, the second component of network pharmacology is engaged: the design or search for compounds that match the desired footprint; the corollary of this is that such compounds should have minimal impact on undiseased networks (based on binding or pleotropic influence) and impact analysis can be applied to normal cells as part of this process.
The steps required in network pharmacology discovery and de-risking are:
1) Construction of a network data-model of the processes of interest using current biological and pathophysiological data
2) Network analysis to identify high-value nodes
3) Computation of optimally synergetic combinations of targets, where optimality relates to network impact and not only to binding affinity at a single protein “target”
4) Selection of molecules that intervene upon the function of the proteins in the target set to exert the desired effect
a) by searching chemoproteomic databases for existing molecules with the desired binding and pleiotropic footprints, or
b) by generating novel compounds assembled from molbits associated with the specific multivalent footprints desired.
5) Computation of the potential impact of the proposed compounds on normal cells’ networks and selection of an optimal subset of lead molecules yielding maximum efficacy with minimum effects on normal cells.
We will describe several real applications of these principles that have generated compounds, for example in cancer therapy and antibiotics, with potent efficacy and acceptable safety profiles.
<<Back to Short Courses