We have recently reported that the statistical analysis of the frequency distribution of short oligonucleotides within mammalian and viral genomes allows the production of sets of DNA sequences enriched in signals for transcription factors. Such statistical approaches could facilitate the identification of new promoter regions playing a role in the transcriptional regulation of gene expression. In the case of mammalian oligonucleotides, we found that the published set of frequent decamers enriched in transcriptional motifs is not suitable for studies on genes of Homo sapiens and evolutionarily related genomes, because it contains decameric sequences belonging to genomic repeats. We report here that most of the decameric sequences of DNA repeats belong to Alu repeats. Accordingly, we produced a subset of Alu-free frequent decamers. In addition, we eliminated from the subset of Alu-free frequent decamers those that are frequently present within other common human repeats, including (GT)n, (AT)n, (CA)n, (ATT)n, (CAA)n and (GTT)n. The Alu-free (repeats-free) subset of frequent mammalian decamers is enriched in signals for transcription factors and allows the identification of putative signals in genes, such as those coding for plasminogen activator, adenosine deaminase and p53, that contain a large number of Alu-like repeats interspersed within our genomic sequences. The newly generated compilation of frequent decamers described here might be used to locate genomic regions playing functional roles in the expression of genes of Homo sapiens and related primates.

The large body of nucleic acid sequence data now available offers a unique opportunity for the characterization of individual oligonucleotides which may be specific to sequence functional domains. We have prepared algorithms for the study of the frequency distribution of all oligonucleotides of length 2-6 in DNA sequences. We have implemented them in the study of 634 mammalian DNA sequences spanning 1.782 Mb, and have obtained the distribution of the ratio between the observed frequency of oligonucleotides and their expected frequency based on independent nucleotide probabilities. We then studied the distribution of oligonucleotides (or k-tuples) of each length in a subset of 129 complete mammalian genes spanning 0.607 Mb. Eight distinct genomic regions, namely 5'-non-transcribed, first exon, first intron, intermediate exons, intermediate introns, last intron, last exon and 3'-non-transcribed, were considered. We observed that some oligonucleotides show a statistical behaviour and a regional distribution similar to that of known signal sequences. Moreover the frequency distribution of oligonucleotides of length 5 and 6 tends to become bimodal, indicating the existence of a population of very frequent oligonucleotides.