Identification of Sequence Motifs Causing Band Compressions on Human cDNA Sequencing

In order to characterize DNA sequences leadingto band compressionsin an automated dideoxy-DNA sequencing system which uses fluorescentdye primers, we compiled DNA sequences at compression sitesfrom accumulated sequence data of human cDNAs (about 205 kbin total length). The results clearly showed that almost allthe 3'-end regions at the compression sites (> 98%) carriedtwo types of common sequence motifs. The predominant one (about68%) contained a sequence of 5'-Y'GN_1–2AR'-3' (Y' andR': pyrimidine and purine residues capable of base pairing).The remainder (about 32%) carried a hairpin motif with a relativelystable GC-rich stem ( 3 bp) connected by a loop consistingof3 or 4 nucleotides. The occurrence of compressions at thesemotif sites was further confirmed by using synthetic DNAs withrandom sequences (about 58 kb in total length). Since DNA sequencesat compression sites analyzed so far shared either of the typeof motifs in the sequencing system employed here, it was possibleto predict the nucleotide residue to be located at a compressionsite by carefully checking the sequence preceding the site.