Mutual information reveals a homonucleotide bias at 6 bp distance in promoters

Abstract

Background: Coding DNA in eukaryotes shows a prominent periodicity at 3-base pair (bp) because of the nonuniform use of bases at the three codon positions. This periodicity can also be interpreted as a correlation between bases at a distance of 3-bp, which in turn implies enhanced predictability at that distance as well as at integer multiples of 3 bp. This study examined possible similar dependencies between nucleotide positions in human promoter region sequences. It is possible that such noncoding sequences might show enhanced predictability at various distances other than 3 bps because of the ubiquitousness of repetitive patterns such as Short Tandem Repeats (STR). It is also possible that the 3D conformational requirements of helical DNA could induce a dependency between bps occupying certain relative positions. Mutual Information (MI), an information-theoretic quantity that can be interpreted as a measure of correlation or predictability between two sequence positions, was used to probe promoter sequences in Homo sapiens. Coding DNA sequences were also probed as a methodological control. Results:The MI measure showed a clear 3-bp signal with integer harmonics in coding DNA sequences. This shows that the MI measure is sensitive to a known repetitive signal in coding DNA sequences. MI in promoter sequences fell with increasing distance between two positions, but also showed an unexpected and significant rise at a distance of 6-bp (i.e., between the first and the seventh bases in a sequence). Harmonic intrusion of a 3-bp signal from coding DNA was ruled out as an explanation for this 6-bp signal in promoter DNA. Purposely misaligning the promoter sequences had essentially no effect on the increased correlation at a distance of 6 bp. Conclusions:Human promoter DNA sequences shows enhanced predictability (i.e., increased mutual information) between a given position and a position 6 bp downstream. This signal runs counter to the general fall in mutual information as the distance between positions increases. The increased correlation at a distance of 6 bp does not depend on the precise alignment of the sequences. Possible explanations for this enhanced predictability are discussed.