How to look for different transcripts for a protein?

Question

I am trying to find all the different transcripts for a protein that translate into different isoforms of the protein. However, when I look it up online, there is no clear organized data on different documented isoforms (at least not for the one I am looking for...) (For example, entries at NCBI Protein database do not match with those at ENSEMBL). It would be great if someone can suggest a more systematic way (or previous such studies) to find out the sequences of different isoforms of a protein.

To elaborate, here the link at ENSEMBL database (9 transcripts, of which 5 are protein coding, of which 2 have identical AAs seq) and here's what I found at NCBI Protein search (4 isoform and a 5th one from literature). It is evident that there are discrepancies between the two.

Answer 1

As will be clear from the question and various replies, there are several different repositories for data emerging from the various nucleic acid sequencing projects. The two main gene oriented ones — NCBI and ENSEMBL — are funded by the US government and European goverments, respectively, and share primary sequence data. However they differ in their interpretative focus and the secondary data (often from elsewhere) that they present in combination with the nucleic acid sequences. There are other valuable webs services that are more limited in scope — the UCSC browser (mentioned by Maximillian Press) providing graphical views of several reference genomes. The main protein-oriented repository is UniProt, as @Michael_A mentioned.

The poster is interested in the transcripts for a particular gene (Transmembrane protease serine 3), and the NCBI and ENSEMBL facilities he consulted were, in my opinion, the most appropriate. As he found different transcripts presented by the two facilities, he might follow the suggestion of consulting UniProt, but it transpires that the set of isoforms it presents differ again. (The UCSC browser displays the ENSEMBL set which originate from the GENCODE project.) This is summarized in the table below — three common isoforms, but ones absent from other data repositories.

NCBI	ENSEMBL	UniProt	Size (aa)
NM_024022.4, NP_076927.1 (iso 1)	TMPRSS3-203	P57727 (-1)	454 aa
NM_032404.3, NP_115780.1 (iso 2)	—	P57727-2	327 aa
NM_032405.2, NP_115781.1 (iso 3)	TMPRSS3-201	P57727-3	344 aa
NM_001256317.3, NP_001243246 (iso 4)	TMPRSS3-208	P57727-5	453 aa
—	TMPRSS3-202	—	451 aa
—	—	P57727-4	293 aa
AAT66641.1 (iso 5)	—	P57727-6	538 aa
[NM_024022.3 (= NM_024022.4)]	[TMPRSS3-209 (= 208)]	—	[duplicate]

Why could cause this conflict, and what can one do to resolve it? Transcript data may be obtained for individual genes in specific studies, or in large projects like ENCODE. Whereas I imagine the commonalities probably arise from the larger projects, it seems likely that the smaller ones are represented in the individual unique cases. The reason that this might be so is that the pattern of differential splicing often differs between tissues and at different developmental stages, so that such transcripts might be missed in whole body studies or in studies in which a tissue was not included. In this case uneven coverage of tissues involved in hearing could possibly be the problem. Mistakes can also arise: this is presumably the reason for the two identical transcripts being listed. And it is difficult for NCBI to co-ordinate different submissions — hence the strange situation of the missing transcript map for ‘isoform 5’.

So there can be no general answer as to which data repository is the “best” or the one to “believe”. One has to find where the evidence is for a particular transcript, i.e. identify the original papers, read them, and make one’s own decisions as a scientist.

Answer 2

I would suggest using the UCSC genome browser for the organism in question (assuming that they have your organism loaded), and use either the GUI or otherwise to pull out the transcripts you are interested in after searching the gene. For example, here is a view of transcripts for the CFTR gene in humans.

You can click on the transcripts in the cartoon, leading you to a page such as this that has a lot of information.

Answer 3

As you are interested in protein coding transcripts it would be worth using a protein database. You can search for curated protein isoforms in Uniprot by limiting your search to "UniProtKB" + the species. Searching for "TMPRSS3" shows 6 protein isoforms. The cross-references section of the page provides links to the RNA transcripts for each isoform in a various DBs. A quick scan confirms that there are direct links to transcripts for 4 of the isoforms; if you check a little closer it should also provide a direct link to the other 2.

Answer 4

I think people more or less expect to do BLAST searches, and also to check the literature when it's available, when they really hope to find all potential splice forms that are currently known. Deciding when a transcript is genuine mRNA rather than partially spliced RNA from the nucleus can be tricky, and may depend on the specifics of how the RNA was isolated (e.g. polysomes vs. whole tissue homogenate; though the difference, while real, seems to be less than expected). Also, there's a philosophical problem: some of the transcripts from your first link barely overlap with one another, and I don't think it's uncommon to look at a stretch of genome with exons ABCDE where AC, DE, and BCDE are all possible splice products - are they "the same" gene?