zeEarly biochemical experiments measuring nearest neighbor frequencies
established that the set of dinucleotide relative abundance values (d
inucleotide biases) is a remarkably stable property of the DNA of an o
rganism. Analyses of currently available genomic sequence data have ex
tended these earlier results, showing that the dinucleotide biases eva
luated for successive 50 kb segments of a genome are significantly mor
e similar to each other than to those of sequences from more distant o
rganisms. From this perspective, the set of dinucleotide biases consti
tutes a 'genomic signature' that can discriminate sequences from diffe
rent organisms. The dinucleotide biases appear to reflect species-spec
ific properties of DNA stacking energies, modification, replication, a
nd repair mechanisms. The genomic signature is useful for detecting pa
thogenicity islands in bacterial genomes.