Neighbor Joining And Leaf Status

The Neighbor Joining Algorithm is among the most fundamental algorithmic results in computational biology. However, its definition and correctness proof are not straightforward. In particular, ‘‘the question ‘‘what does the NJ method seek to do?’’ has until recently proved somewhat elusive’’ [Gascuel & Steel, 2006]. While a rigorous mathematical analysis is now available, it is still considered somewhat hard to follow and its proof tedious at best. In this work, we present an alternative interpretation of the goal of the Neighbor Joining algorithm by proving that it chooses to merge the two taxa u and v that maximize the ‘‘leaf-status’’, that is, the sum of distances of all leaves to the unique u-v-path.