The analysis of intron data and their use in the detection of short signals

In order to examine whether certain short DNA sequences (putative splice signals) occurred in a certain region of an intron more often than would be expected by chance, intron data were examined to see what structure they took. There were significant departures from equal nucleotide frequency, and successive nucleotides clearly did not occur independently in the rat and mouse introns examined. The nonindependence was mainly due to a CG shortage and a less marked TA shortage. However the pairwise frequencies explained almost all the variability in triplet frequencies in the data and so the data could be approximately modeled by using nucleotide frequencies conditional on what the previous nucleotide was. Some coding DNA was also examined and the pairs in second and third positions, and third and first positions in a codon, showed similar departures from independence to those of the intron data. Using the probability model derived for intron data, expected frequencies of putative signals were derived and compared with the observed frequencies.