A novel bioinformatics strategy for unveiling hidden characteristics in genome sequences and searching in silico for genetic signal sequences

摘要

Novel bioinformatic tools are needed for comprehensive analyses of massive amounts of available genome DNA sequences. An unsupervised neural network algorithm, self-organizing map (SOM), is an effective tool for clustering and visualizing high-dimensional complex data on a single map. We generated SOMs for tri-, terra-, and pentanucleotide frequencies in 300,000 10-kb sequences from 13 eukaryotes for which almost complete genomic sequences are available (a total of 3 Gb). SOM recognized in most 10-kb sequences species-specific characteristics (key combinations of oligonucleotide frequencies), permittingrnspecies-specific classification of sequences without any information regarding the species. Because the classification power is very high, SOM is an efficient and powerful tool for extracting a wide range of genomic information. SOM constructed with oligonucleotide frequencies in 10-kb sequences from 2.8 Gb of human sequences identified oligonucleotides occurring with frequencies characteristically biased from random occurrence predicted from the mononucleotide composition; 10-kb sequences rich in these oligonucleotides were self-organized on a map. Because these oligonucleotides often corresponded to genetic signals or their constituent elements, we propose an in silico method that should be useful for identification of genetic signal sequences in genomes for which large amounts of sequence data are available but additional experimental data are lacking.