Is anybody aware of any benchmarks of quality and quantity of genes predicted by genie and ensmbl from December freeze data?
Response: Ensembl could be consulted directly for benchmarks on their predicted genes. The number and details for both prediction systems are accessible from the mysql text browser if you set the “select a database” to hg6 (Dec 00 build). The proper field names are not given in the list (being fixed) so you would need to watch the urls and make the appropriate changes (1,2). Using the data for the two prediction systems, you could fairly easily compare them genomewide to each other and to known genes. But looking at chr9:106342105-106604031 for the Dec 00 assembly, you can see right away that known genes, ensembl genes, and genie genes do not agree with each other completely, nor for that matter seem to be aware of the full length mRNA set. Indeed, for the gene ABCA1, there are at least 3 mRNAs supporting an additional upstream exon. Thus it is not clear where one obtains a gold standard for measuring gene prediction quality A reference set might be hand curated out of recent full text
Related Questions
- Do you offer any benchmarks of quality and quantity of known and predicted genes shown in the assembly tracks from RefGene, Acembly, Ensembl, Genscan, Fgenesh++, and TIGR Gene Index?"
- What happens if I don follow CEF guidelines, use lower quality product, or use a lesser quantity than approved by CEF?
- Is anybody aware of any benchmarks of quality and quantity of genes predicted by genie and ensmbl from December freeze data?
