The contribution of data analysis in the study of Lactococcus lactis biodiversity.

We have used the Partial Least Squares (or Projection on Latent Structures) regression (PLS) to describe relationships between molecular profiles obta ined by Randomly Amplified Polymorphic DNA (RAPD), considered as the X vari ables, and phenotypes, considered as the Y variables. In our examples, we s how that RAPD profiles can be correlated with strain phenotypes (e.g., lact is or cremoris phenotypes, diacetyl production capacity, lytic bacteriophag e group classification) of lactococcal strains. The PLS regression models w hich describe these relationships, as well as the data sets used for their calculation, can be used to predict phenotypes of unknown strains using onl y their RAPD profiles. This kind of analysis opens attractive perspectives for improving the description and for the valorisation of the biodiversity of our microbiological ressources.