Figure 8

Normalization decreases coverage of highly abundant genes, but does not change the GO term distribution of contigs. In both samples, most contigs are composed of <10² reads. However, the non-normalized sample (A) contains contigs with many more reads per contigs than the normalized sample (B). In other words, normalization preferentially decreases the number of reads of those contigs with the most reads. (C) GO term distributions do not differ dramatically between pyrosequenced libraries of N versus NN cDNA. However, see Additional file 6 for exceptions. Column heights reflect the percentage of annotated sequences in each assembly that mapped to a given GO term. Note that the GO terms shown represent the results of mapping the N and NN reads against the complete assembly, rather than those obtained via independent assemblies of N and NN reads.