Long range cross correlations between nucleotide triplets in human chromosomes

Long range autocorrelations of nucleotide pairs in human chromosomes have been studied widely in the literature. Long-range correlations differ from usual correlations in that they indicate underlying fractal nature. Despite the wide spread acceptance of this observation for nucleotide pairs, its biological significance has not been completely explored. Motivated by the fact that the basic unit for aminoacid synthesis are codons, composed of three nucleotides, in this paper, we calculate the correlations between pairs of nucleotide triplets. We utilise an approximation to the mutual information function suggested by Beier, which is approximately equal to correlation. We underline that we do not consider only autocorrelations but also crosscorrelations. We study the human choromosomes 22 and 23. We demonstrate that among 4096 such triplet pairs, 23 show long range dependence. Considering a Gaussian distribution for the crosscorrelation of such pairs, these 23 pairs stand out at 3 sigma distance from the mean. The list of these pairs are: AAA-AAT, AAA-ATA, AAA-ATT, AAA-TAA, AAA-TAT, AAA-TTT, AAA-ATA, AAT-ATT, AAT-TAT, AAT-TTT, ATA-ATT, ATA-TAA, ATA-TAT, ATA-TTA, ATA-TTT, ATT-TAA, ATT-TAT, ATT-TTT, TAA-TTT, TAA-TTA, TAT-TTT, TTA-TTT. The biological significance of this result is yet not clear and we believe that this conference would be an ideal platform for discussion on the interpretation of these observations.