As the number of sequenced genomes increases, the single string is rapidly becoming an ill-suited approximation, discarding accessory genes, rearrangements and repeated regions. I will discuss genome representations which are well suited to statistical analysis when genomes cannot be aligned or even assembled.