The evolutionary reason for the increase in gene length from archaea to prokaryotes to eukaryotes observed in large scale genome sequencing efforts has been unclear. We propose here that the increasing complexity of protein-protein interactions has driven the selection of longer proteins, as longer proteins are more able to distinguish among a larger number of distinct interactions due to their greater average surface area. Annotated protein sequences available from the SWISS-PROT database were analyzed for thirteen eukaryotes, eight bacteria, and two archaea species. The number of subcellular locations to which each protein is associated is used as a measure of the number of interactions to which a protein participates. Two databases of yeast protein-protein interactions were used as another measure of the number of interactions to which each \emph{S. cerevisiae} protein participates. Protein length is shown to correlate with both number of subcellular locations to which a protein is associated and number of interactions as measured by yeast two-hybrid experiments. Protein length is also shown to correlate with the probability that the protein is encoded by an essential gene. Interestingly, average protein length and number of subcellular locations are not significantly different between all human proteins and protein targets of known, marketed drugs. Increased protein length appears to be a significant mechanism by which the increasing complexity of protein-protein interaction networks is accommodated within the natural evolution of species. Consideration of protein length may be a valuable tool in drug design, one that predicts different strategies for inhibiting interactions in aberrant and normal pathways.

13 pages, 5 figures, 2 tables, to appear in Physica A