Computing genetic similarity coefficients from RAPD data: the effects of PCR artifacts.

Random amplified polymorphic DNA (RAPD) markers have been used for many types of genetic analyses, including genome mapping, genotype fingerprinting, phylogeny reconstruction, and measuring genetic similarities. They suffer from one potential limitation, however, because the PCR that is used to produce informative amplification products often produces artifactual products as well. Optimization of PCR protocols to eliminate artifactual bands completely is often too costly or too time-consuming to be practical. Other methods for handling RAPD artifacts, such as deleting inconsistent or faint bands or using only those bands that are reproducible, introduce false negatives into the data. Simply ignoring artifacts and using all bands introduces false positives. When RAPD data are used to compute genetic similarity coefficients, such artifacts can cause significant bias in the estimation. The three coefficients most widely used with RAPD data, the simple matching coefficient, Jaccard's coefficient and Nei and Li's coefficient, differ in the amount of bias produced by a given level of artifactual bands. The simple matching coefficient and Nei and Li's coefficient always exhibit less percent bias than Jaccard's coefficient. For closely related organisms, Nei and Li's coefficient displays less percent bias than the simple matching coefficient. If new DNA samples possessing RAPD markers not present in the previously analyzed samples are added to a study, values of the simple matching coefficient will need to be computed for all samples, not just the new ones. Jaccard's and Nei and Li's coefficients, however, will not need to be recomputed. Furthermore, only Nei and Li's coefficient has a direct biological meaning (it is an estimate of the expected proportion of amplified fragments shared by two samples because they were inherited from a common ancestor). On the basis of these results, Nei and Li's coefficient is recommended for routine computation of genetic similarities using RAPD data, particularly if PCR artifacts are present.