UPGMA ( U nweighted P air G roup M ethod with has rithmetic mean ) is the name of a algorithm intended for the construction of a Phylogenetic tree. This method allows the transformation of a matrix of distances (between various organizations, populations, or sequences of Nucléotide S) into a tree enraciné.

The matrix provides the whole of the distances between all the pairs of elements. The algorithm functions by successive iterations, which reduce the size of the matrix gradually. Each iteration sees the regrouping of the two remaining elements separated by the shortest distance: these elements are associated in the tree, and are replaced by an element “consensus”. The new distances between this element consensus and the remaining elements in the matrix are recomputed by the arithmetic Mean of the two gathered elements.

This simple method and rapid present however many skews. In particular, it supposes that the speed of evolution is constant in all the branches. By consequence, if an “internal” branch evolves/moves much more quickly than all the others, it will be attached to the remainder of the tree only at the last stage and will be outside the tree (the phenomenon is similar to the Attraction of the long branches).

The defects of the UPGMA are not such as the algorithm does not have any more but one interest historical. It was indeed replaced by more advanced methods since then (like the Neighbor Joining or the Parcimonie initially, then techniques of Maximum of probability or algorithms bayesiens used today in Phylogénie).