Illustration of the major benefit of de novo clustering. A real cDNA is shown as a brown bar, and short reads originating in it are merged into nonredundant reads with unique sequences presented by gray bars. (A) The contrast densities of the gray bars are proportional to their frequencies. The reference genome is shown as a long arrow flagged with the corresponding locus of the cDNA by a brown bar. (B) Alignments of best hits are highlighted by blue dashed lines. Red dots emphasize base positions at which the reads disagree with the original cDNA sequence. The direct alignment includes correct alignments, as well as some short reads with multiple best hits, as illustrated by the leftmost read. Some reads fail in alignment because the sequencing errors are too numerous, as shown by the aslant bar, and some are aligned to false-positive positions. (C) These short reads are organized into the tree by the proposed de novo clustering before alignment with the genome. The root indicates the representative sequence of the cluster; this is the darkest, most abundant read denoted with an asterisk. In the tree, parent–child relationships are depicted by dashed lines.
