support for long range exon combinations (15). RefSeq curators mitigate against false positives by selectively incorporating only high quality datasets for consideration into our genome annotation pipeline and into the manual annotation process. RefSeq curators visualize transcript alignments, variation data, and filtered RNA-Seq data in customized displays within an in-house alignment tool incorporated into the NCBI Genome Workbench platform (http://www.ncbi.nlm.nih.gov/tools/gbench/). Curation of human genes utilizes analyzed RNA-Seq reads from the Illumina BodyMap 2 (BioProject: PRJEB2445) and Human Protein Atlas projects (BioProject: PRJEB4337) (16). Additionally curators use promoter-associated histone modification marks such as H3K4me3 from the NIH Roadmap Epigenomic Mapping Consortium (REMC; (17) and the ENCODE (Encyclopedia of DNA Elements) project (18) to verify the presence of an active promoter. RefSeq curators also evaluate polyA-seq data to affirm 3′ completeness of transcripts lacking a polyA tail (19). Additional data types, including PhyloCSF (20), CpGIslands, RepeatMasker (21) and Cap analysis of gene expression (CAGE) data (22), are sometimes used as additional support.
