Periodicities of dinucleotide self-information values in phi X174 DNA.

The natural DNA sequence of bacteriophage, phi X174, when analyzed as a "text" of dinucleotides, is shown to display an easily detectable degree of non-randomness by the distribution of values of dinucleotide self-information along the sequence. Self-information corresponding to occurrences of dinucleotides separated by a single nucleotide is found to be somewhat higher than the values which precede or follow it for every third nucleotide position along the sequence. Consequently autocorrelation coefficients of these values display a strong periodicity and harmonic analysis of the values shows a spike at a value of 3. Self-information autocorrelation periodicity is used as a test of the effect of randomizing portions of the sequence. Any one or two of the three nucleotides in each triplet of the sequence can be chosen at random without losing dinucleotide self-information periodicity except when both the 1st and 3rd nucleotide of all of the triplets in the major phi X174 protein reading frame are randomized. Periodicity is also lost when sequences are generated by randomizing triplets. Autocorrelation and harmonic analysis also indicate other less marked periodic features of dinucleotide self-information values of the nature sequence; non-random features are suggested at periods of 12, 20 and 24 nucleotides.