The maximum information principle and the evolution of nucleotide sequences.

The probability distributions of bases in nucleotide sequences are deduced from the maximum information principle by maximizing the entropy (due to random mutation of bases) under certain constraints (Markovian entropy, G + C content, etc, due to selection). Two formulations are given with respect to different selective constraints. The deviations of theoretical distributions from experimental data are lower than 10% for most sequences. It is shown that the Lagrange multipliers change from species to species systematically--i.e. selective constraints correlate with evolution.