An improved procedure for clustering and assembly of large transcriptome data