Many gene superfamilies have hundreds or thousands of members and hence pose a significant challenge when performing a large-scale phylogenetic analysis. Derivation of the most accurate alignment possible and inference of evolutionary relationships (with an appropriate measure of confidence) are significant "bottlenecks" in the process. A generally applicable strategy is outlined for identifying and aligning sequences, performing simple analysis of the resulting alignment, and inferring evolutionary relationships. Reference is made to the serpin superfamily. The 'partition cluster' method, a relatively rapid technique for extracting underlying associations from phylogenetic bootstrap trees, is also presented.