We propose a new distance-based clustering method, triplet clustering algorithm (STC), to reconstruct phylogenies. The main idea is the introduction of a natural definition of so-called k-representative sets. Based on k-representative sets, shortest triplets are reconstructed and serve as building blocks for the STC algorithm to agglomerate sequences for tree reconstruction in O(n^2) time for n sequences. Simulations with 500, 1000 and 5000 sequences data sets show that STC gives better topological accuracy than other methods tested.
