Phylogenetic inference from binary sequences reduced by primary DNA sequences

Given a bi-classification of nucleotides, we can obtain a reduced binary sequence of a primary DNA sequence. This binary sequence will undoubtedly retain some biological information and lose the rest. Here we want to know what kind of and how much biological information an individual binary sequence carries. Three classifications of nucleotides are explored in the present article. Phylogenetic trees are built from these binary sequences by the Neighbor-Joining (NJ) method, with evolutionary distance evaluated on the basis of a symbolic sequence complexity. We find that, for all data sets studied, binary sequences reduced by the purine/pyrimidine classification give reliable phylogeny (almost the same as that from the primary sequences), while the other two result in discrepancies at different levels. Some possible reasons and a simple model of sequence evolutionary are introduced to interpret this phenomenon.