Analyzing large data sets: rbcL 500 revisited.

In 1993, Mark Chase and 41 coauthors published phylogenetic analyses of two very large data sets of nucleotide sequences of the chloroplast gene rbcL, which encodes the large subunit of ribulose 1,5-bisphosphate carboxylase. Their paper was important for several reasons. These analyses were (and still are) among the largest ever attempted using parsimony. The assembly of such a large number of sequences clearly demonstrated a high level of cooperation on the part of the botanical systematics community. Furthermore, a number of important new hypotheses regarding seed plant phylogeny emerged from this study, and it has helped to orient many subsequent phylogenetic analyses. Increasingly, the Chase et al. trees are being used in quantitative comparative analyses (e.g., Barraclough et al., 1996; also see Donoghue and Ackerly, 1996, and associated papers). We reanalyzed one of the Chase et al. data sets for two reasons. First, we wanted to explore the general methodological and theoretical issues raised by very large data sets. It is critical that these issues be addressed now because the number of large data sets is increasing rapidly. Second, in view of its importance, we wanted to discover the effects of long search times and alternative search strategies on this data