We developed computer-based methods for constructing a nonredundant mouse full-length cDNA library. Our cDNA library construction process comprises assessment of library quality, sequencing the 3Ј ends of inserts and clustering, and completing a re-array to generate a nonredundant library from a redundant one. After the cDNA libraries are generated, we sequence the 5Ј ends of the inserts to check the quality of the library; then we determine the sequencing priority of each library. Selected libraries undergo large-scale sequencing of the 3Ј ends of the inserts and clustering of the tag sequences. After clustering, the nonredundant library is constructed from the original libraries, which have redundant clones. All libraries, plates, clones, sequences, and clusters are uniquely identified, and all information is saved in the database according to this identifier. At press time, our system has been in place for the past two years; we have clustered 939,725 3Ј end sequences into 127,385 groups from 227 cDNA libraries/sublibraries (see http://genome.gse.riken.go.jp/).[The sequence data described in this paper have been submitted to the DDBJ data library under accession nos. AV00011-AV175734, AV204013-AV382295, and BB561685-BB609425.]The collection of full-length genes requires libraries with a high content of full-length cDNA inserts, largescale sequencing, library assessment, and high-speed sequence clustering . Here, we focus on computational methods, such as newly developed computer programs, because our experimental methods have been published previously (Carninci et al.