Under-representation of PolyA/PolyT tailed ESTs in human ESTdb: an obstacle to alternative polyadenylation inference

Alternative polyadenylation is a key regulatory process which affects the 3' end formation of variants of the same transcription unit, thus altering gene expression pattern, and transcripts' cellular behaviour and characteristics. The common methodology for computational analysis of alternative polyadenylation signal utilization is based on EST data, specifically on PolyA/PolyT tailed ESTs. Studying the human ESTs dataset we detected a significant underrepresentation of PolyA/PolyT tailed ESTs, constituting only 10% of most libraries. Consequently, more than 50% of false-negative events are revealed in the analysis of alternatively polyadenylated variants' expression. We therefore argue that the ratios of PolyA/PolyT tailed ESTs, as represented in the human EST database, do not reflect the truepicture of 3' end variants formation of a given physiological situation. Thus the EST database should not be considered a reliable source for alternative polyadenylation signal usage inference.