Growth of microbial genomes by short segmental duplications

A DNA sequence can be analyzed as a text of four letters by counting the times each word in the set of k-letter words occurs in the text. If the text is random and long enough, then the frequencies of word occurrence are expected to obey a Poisson distribution. Examination of complete microbial genomes shows that for k less than 9, the distribution has a width many times the width of a Poisson distribution 42, 24, 9 and 3.2 times, for k being 2, 4, 6 and 8, respectively. The cause of this phenomenon is not known. Here we propose a simple biologically plausible model for the growth of genomes to explain it: the genome first grows randomly to a length much shorter than its final length, thereafter mainly grows by random segmental duplication. We show that using an initial length of 1000 bases (1 kb) and duplicated segments with lengths averaging 25 b, one can generate a model sequence the size of microbial genomes of the order of 1 Mb that exhibits genomic statistical characteristics.