In the past, the construction of complementary DNA (cDNA) libraries has served as a key technology for the discovery of biologically interesting genes (1). For mammalian genes, this is true even in the postgenome sequencing era. The prediction of protein-coding sequences solely from the genomic sequence is stilldifficult because protein-coding sequences are divided into multiple small pieces by introns. Therefore, currently, cDNA libraries are frequently used to complement genomic information for the comprehensive analysis of cDNAs, rather than the discovery of particular genes. For this purpose, a large number of randomly sampled cDNA clones are usually analyzed using sequencing, whereas cDNA clones are subjected to characterization after screening processes for the purpose of gene discovery. In practice, comprehensive analysis differs from that of conventional gene discovery in that comprehensive analysis is most frequently carried out on a single-clone-for-single-gene basis, whereas gene discovery is done through the isolation of multiple cDNA clones for a single gene. Thus, current requirements for high-quality cDNA libraries suitable for comprehensive cDNA analysis are (1) high rate of full-length and authentic copies of mRNAs, (2) low population bias of cDNA clones, (3) high complexity, (4) low occurrence rate of artificial clones such as chimera, and (5) low occurrence rate of immature mRNA copies. As for requirement (1), many groups have made a tremendous effort to date and have used their methods in practice (2–7).