annotations, where both processes result in the same annotation. For the analyzed subset of GENCODE, 38% of the transcripts have good full-length support and an additional 39% have a lower level. The different distribution of support levels between the annotations is consistent with the underlying approaches. The merged subset is heavily weighted toward well-supported transcripts. The higher fraction of manual-only annotations supported only by ESTs is due to Ensembl not directly using ESTs in their pipeline.
