Amplified fragment length polymorphism

Although AFLP should not be used as an acronym, it is commonly referred to as "Amplified fragment length polymorphism" (borrowing the term "Amplification fragment length polymorphism" used in DNA amplification fingerprinting). However, the resulting data are not scored as length polymorphisms, but instead as presence-absence polymorphisms.

AFLP-PCR is a highly sensitive method for detecting polymorphisms in DNA. The technique was originally described by Vos and Zabeau in 1993. In detail, the procedure of this technique is divided into three steps:

1. Digestion of total cellular DNA with one or more restriction enzymes and ligation of restriction half-site specific adaptors to all restriction fragments.
2. Selective amplification of some of these fragments with two PCR primers that have corresponding adaptor and restriction site specific sequences.
3. Electrophoretic separation of amplicons on a gel matrix, followed by visualisation of the band pattern.

Analysis

As mentioned earlier, the results of AFLP is scored as several sequences of presence-absence polymorphisms. The "sequences" can then be compared and hierarchical clustering can be used to produce a dendrogram.

• With methods such as UPGMA (commonly used for AFLP) and neighbor joining, a distance matrix is required. There are several ideas on how a distance matrix can be made:
  • Hill et al. (1996) uses the formula of Nei and Li (1979), which scores the presence or absence at each discrete length (0/1 marker sequence).
  • Not all gels are run to the point where lengths can be resolved in increments of 1 base (examples include the two images in the section), and even if they are distinguishable, counting bands is error-prone. Since 1999, studies deal with this issue by instead considering each result a sequence of density levels (like a one-pixel-tall picture). These sequences are compared to each other using Pearson correlation. This also has the advantage of making use of different density levels.
• Two separate Bayesian methods, one from MrBayes, the other from Luo et al. (2007), first translate the result into 0/1 marker sequences.
• One Bayesian method translates density-sequenecs into population structure parameters.

The issue with 0/1 marker sequences is that they are inherited in a dominant way, but most statistical methods used in ecology require codominant data. The density-sequence representation of the result of AFLP is suitable for dealing with this issue, so long as the change in density is due to actual change in the relative amount of nucleic acids of a certain size and not experimental error.