Schematic presentation of the data-extraction method. Sequences are extracted locally around potential ESE motifs and are then declined or accepted as positive examples depending on whether they fulfill certain criteria. First of all, a potential positive example has to be close to a splice site. Secondly, the exon from which the positive example is extracted has to be a weak exon and thirdly, the region in which the positive example is located has to be single-stranded. Each training sequence has a length of 200 bases.