One reason why you cannot find the sequence might be that it is NOT stored in the database under the definition name "insulin receptor substrate 2" that you have looked for but under another name, e.g. "sequence similar to isr-1" or "sequence found by screening with moAb to isr-1".
If you do not have the original paper of the publication with the accession number at hand, you could try to search e.g. at EMBNETs WWW-search site.
Here in Vienna, I use the search engine of the local EMBnet node that can be reached at "http://vienna.at.embnet.org/srs/srsc/": A nice site pops up. Klick at "search sequence libraries". Then select that you want to search "EMBL" as well as "GENBANK" and "EMNEW". In the fields below, chose "definition" instead of "all text" in the first three lines, and enter "insulin" in the first line, "receptor" in the second line, and "substrate" in the third line. In the last line, change "all text" to "organism", and enter "human". The queries are combined with "AND" (preset, is OK). Then, you should mark what fields you want to have listed, e.g. mark everything from "ID" to "comment" (or to "link"). Then klick "DO QUERY".
One important fact to remember is that GENBANK and EMBL databases are ALMOST identical, though not to 100%. Because of this, EMBnet provides only those sequences in the local GENBANK database mirror that are not already present in EMBL. (It would not make sense to store redundant information twice!) In other words, if you search in Genbank only, you practically do a search "genbank NOT embl"!!! Therefore, always chose genbank as well as "embl" when searching at an EMBnet node!
When I did the search mentioned above, 11 results came up. The ACs are:
The results contain a few sequences that are described as "clone similar to insulin receptor substrate 1" etc.
Hopefully, you can find the one among it that you think is called now "isr-2".
Hope that helps
peter.pichler at vm.akh-wien.ac.at