DNA–DNA hybridization generally refers to a molecular biology technique that measures the degree of genetic similarity between pools of DNA sequences. It is usually used to determine the genetic distance between two organisms. This has been used extensively in phylogeny and taxonomy.

The DNA of one organism is labelled, then mixed with the unlabelled DNA to be compared against. The mixture is incubated to allow DNA strands to dissociate and then cooled to form renewed hybrid double-stranded DNA. Hybridized sequences with a high degree of similarity will bind more firmly, and require more energy to separate them: i.e. they separate when heated at a higher temperature than dissimilar sequences, a process known as "DNA melting".

To assess the melting profile of the hybridized DNA, the double-stranded DNA is bound to a column and the mixture is heated in small steps. At each step, the column is washed; sequences that melt become single-stranded and wash off the column. The temperatures at which labelled DNA comes off the column reflects the amount of similarity between sequences (and the self-hybridization sample serves as a control). These results are combined to determine the degree of genetic similarity between organisms.

When several species are compared, similarity values allow organisms to be arranged in a phylogenetic tree; it is therefore one possible approach to carrying out molecular systematics.

DNA–DNA hybridization was once used as a primary method to distinguish bacterial species; a similarity value greater than 70% is described as indicating that the compared strains belonged to distinct species.^[1][2][3] In 2014, a threshold of 79% similarity has been suggested to separate bacterial subspecies.^[4]

Charles Sibley and Jon Ahlquist, pioneers of the technique, used DNA–DNA hybridization to examine the phylogenetic relationships of avians (the Sibley–Ahlquist taxonomy) and primates.^[5][6]

Critics argue that the technique is inaccurate for comparison of closely related species, as any attempt to measure differences between orthologous sequences between organisms is overwhelmed by the hybridization of paralogous sequences within an organism's genome.^[7] DNA sequencing and computational comparisons of sequences is now generally the method for determining genetic distance, although the technique is still used in microbiology to help identify bacteria.^[8]

The modern approach is to carry out DNA–DNA hybridization in silico using completely or partially sequenced genomes.^[9] The GGDC ^[10] developed at DSMZ is the most accurate known tool for calculating DDH-analogous values.^[9] Among other algorithmic improvements, it solves the problem with paralogous sequences by carefully filtering them from the matches between the two genome sequences.