What is the stabiity of Ensembl IDs at the gene, protein and transcript levels ?
I know human is pretty solid (but still not completely 1:1:1:1 with HGNC, Swiss-Prot and Entrez Gene) and there is a push to close mouse and Zebrafish via Havana
However I guess most of the other assemblies are provisional to some extent and the gene IDs therfore are difficult to lock down across re-builds, that will also cause some churn in the transcripts and ORFs
Is there any data on this ? The reason for asking is in regard to citing them and/or including raw sequence as supplimentary data