 |
Center for Nonlinear Studies |
The dynamics of complex biochemical networks typically depend on many parameters (e.g. reaction rate constants). I show that mathematical models of such networks exhibit a universal "sloppy" pattern of parameter sensitivities; their dynamics are exponentially more sensitive to changes in some combinations of parameters than others. For model builders this suggests that predictions will be much more efficiently constrained by fitting parameters than by directly measuring them. I also briefly explore the evolutionary consequences of sloppiness in the context of Fisher's geometric model, showing that sloppiness has little affect on the first step in an adaptive walk, but that it may substantially slow the long-term pace of adaptation.
Host: Byron Goldstein, T-10