1. AceView Danielle and Jean Thierry-Mieg NCBI http://www.aceview.org = global annotation of the whole human genome ● Restricted to the Gencode Regions ● Reinforced filters to obtain Havana density

2. AceView Strategy ● Align all (4.3M) human mRNAs and ESTs ● Refine the introns by co alignment ● Cluster into transcripts and genes ● Annotate and ignore the suspect clones ● Annotate the resulting proteins ● Do not use a priori splice consensus ● Do not mask the genome or the cDNAs

3. Gencode adaptation ● Ignore 5' 3' pairs (they were creating gaps) ● Reassess manually all the atypical introns and ask the code to ignore the corresponding dubious clones ● Ignore all the NMs ● Remove the pseudo genes To ignore a feature is the only manual operation authorized in the AceView annotation interface

4. About the introns boundaries 98.5% are gt-ag 1% are gc-ag Most other are dubious. If the suspect intron is supported by less than N clones and its boundaries fall within exons supported by some other clone, we flag the N clones as 'suspected internal deletion' and ignore this intron For Gencode we chose N = 3 some of these suspect introns were kept in Havana

5. New definition of a gene 2 transcripts belong to the same gene if they iterativelly share at least one donor or one acceptor site ● Split genes touching by their UTR ● Ignore stupid EST bridges ● Shed unspliced mRNAs/EST ● Respect true complex loci

6. Comparison AceView/Havana Good agreement: on number of genes and transcripts. The reinforced filters will be used in the next global AceView release (later this month). Havana is missing some Aceview score=1 transcripts fully supported in genbank: mostly cassette exons i.e.: BI910947, CF593648 A few dubious introns were kept in Havana