Figure 1

Power-law distribution of family clusters in MYRbase with maximal 40% sequence identity and a minimal size of 3 (to exclude false positives). The inset table gives the values and correlation coefficients for different minimal cluster sizes. The power function distribution (without shift along the argument axis x), the Pareto distribution and the Zipf's law (if the rank is approximated by the argument x) have the common analytical form y = a · x^b. Such distributions generally occur as the limit distribution of a multiplicative stochastic process with a lower boundary constraint (here, minimal cluster size). The common phenomenological form does not imply a unified mechanism for generating samples obeying these distribution functions. We suggest interpreting cluster size in terms of time for evolutionary divergence within the cluster (see text).