1.3. Definition of the parameter boundaries for GSA Setting the boundaries of the parameter space for GSA for large scale models represents a distinct task, as on the one hand, they should be relatively wide to justify the globality of the analysis, but on the other hand the boundaries should be reasonably Z-VAD-FMK supplier narrow due to the limitations imposed by the resulting computational time and available CPU resources. Since our GSA implementation is specifically directed towards identification of appropriate drug targets and cancer-related biomarkers within signalling networks, the parameter ranges should be able to incorporate potential
effects of drugs and genetic modifications on the level of protein activities. In our analysis we assumed that up to a 10-fold Libraries reduction in parameter value could imitate an efficient suppression of the protein activity by an anti-cancer drug. It’s this website worth noting, that it is difficult to predict the real extent of the
inhibition of the protein activity by targeted drugs in vivo, since it depends on many factors – drug transformations within the body, efficiency of drug delivery to the target, etc. However, there is a good reason to believe that in vivo drugs cause not more than a 10-fold inhibition of targeted protein activity. For example, in our experiments pertuzumab caused up to 40% inhibition of ErbB3/2 dimer formation ( Faratian et al., 2009b). Recent findings of Gaborit et al. (2011) also confirmed that anti-ErbB2 drugs cause not more than 40–20% of reduction of ErbB2 heterodimerization, when used alone, and up to 70%, when combined with an EGFR inhibitor. These estimates have been made for drugs targeting cellular membrane receptors. For intracellular targets the level of inhibition may be even lower, check due to additional factors, limiting drug availability within the cell (e.g. due to inefficient drug transfer into
the cell). Similarly, we assumed that up to a 10-fold variation of parameter value above and below its nominal value (that in total provides effectively a 100-fold variation) could approximate modification of protein activity by the majority of mutations. For example, a PIK3CA mutation is thought to increase PI3K activity only two-fold (Carson et al., 2008), whereas lipid phosphatase activity of PTEN can differ up to 100-fold between different PTEN mutants, as assessed in (Rodriguez-Escudero et al., 2011). Importantly, in our analysis the parameters are varied within the 10-fold range around the nominal value, thus allowing us to consider many possible levels of protein inhibition/activation, including both weak and strong effects. Thus, for our ErbB2/3 network model the constraints for the majority of kinetic parameters were set to span one order of magnitude above and below the values obtained in one of our best data fits. In some cases the parameter ranges were adjusted to match the order of magnitude of other existing estimates (see Additional File 2 and Table S2).