Clinical trials to increase expression of one or even multiple VEGF ligands in CAD and PAD have not succeeded (Sidebar 1
). For a multicellular physiological mechanism as complex as microvascular remodeling, with multiple cytokine families involved [59
], a more nuanced approach is needed. In particular, quantitation is a core requirement for the design of new therapies. This is where systems biology – a synergistic combination of computational models and quantitative biological experiments – is most important. Without extensive measurement of in vitro and in vivo mRNA, proteins and signaling pathways, as well as anatomical and physiological parameters, computational models may not yield meaningful results; likewise, without the ability to integrate the wealth of experimental data into models, therapeutic design is somewhat blind.
By establishing a solid quantitative foundation, the systems biology approach allows the prediction of the most promising target, and can optimize dosing, timing and even combinatorial delivery of multiple agents. This approach allows the rapid analysis of multiple conditions and in-vivo studies can then establish which set of modeled conditions is most appropriate. By allowing researchers to test many approaches in silico, the success rate of the translation to in vivo drug or gene delivery can be improved.
In addition, we aim to ultimately identify subgroups of the patient population who will respond well to different therapies. Just as trastuzumab (Herceptin) works only for those with HER-2-positive breast cancer, so too the individual variability in ischemic diseases (e.g. baseline VEGF ligand/receptor expression) can be used to design specific therapies for the affected subpopulation. In ischemia, or during exercise training, multiple elements of the VEGF pathway exhibit altered expression [51
]; we can use systems biology to test therapies that mimic these multifactorial changes.
Among the quantitative experimental approaches useful for systems biology, characterization of cytokine networks include microarrays (for expression data), proteomic and phosphoproteomic arrays (for protein concentrations) and mass spectrometry. A subset of the data is incorporated into the computational models as parameters or inputs; the remainder is reserved for comparison to model outputs (validation). The computational models are therefore designed to be molecularly-detailed; that is, they specifically include all the molecules involved to enable comparisons to experimental data. Models may be developed ab initio
or use one of many modeling platforms to create and simulate experimental conditions, including BioNetGen [17
], VCell [66
], SBMLToolbox [42
], ECell [91
] and others.
It is important to note that Systems Biology is not an end in itself; one of the primary goals in creating and using computational models such as those described in this review is to generate new, and most importantly, testable hypotheses to guide scientific progress. As with any model system, including cell culture and pre-clinical animal models, predicted optimal treatments must be extensively tested for safety, efficacy and more. In addition, as hypotheses are tested, both positive and negative outcomes are instructive. Model-building is, in a sense, never complete, and negative outcomes inform additions, revisions and improvements to the model, to then generate improved hypotheses. However, predicted optimal treatments might not be arrived at – or might have taken much longer to generate – without the computational models.