The number of alternative mRNA splice forms that map to human protein coding loci has increased to the point that nearly all proteins have such associated database records. This gives rise to the paradox that the gene build pipeline from the latest Ensembl GRCh38 reference genome assembly indicates 19,919 protein coding loci (which shrinks to 19,022 with HGNC annotation stringency) but 198,002 transcripts (i.e. nearly 10 transcripts per protein). Their is no question that a small number of these alternative splice forms, AS, (plus alternative initiations) have not only been verified to exist as proteins, have some kind of alternative biochemical functions and are also of pharmacological importance . Notwithstanding, compared to the massive transcript profiling that RNAseq now provides routinely, experimentally verifying AS existence at the protein level at large scale is extremely difficult. This is because it can only be done by splice form specific antibodies, western blots detecting different size forms, top down proteomics (i.e. intact mass measurement) or the detection of alternative exon-specific trypic peptides. A recent review  proposes that expanding data sets from the latter approach are consistently detecting only single quantitatively dominant protein isoforms from each locus. The provocative inference is that the vast majority of the 200K odd predicted and/or verified alternative mRNA transcripts are not actually translated into proteins. This can be seen as an interesting methodological detection “gulph” between RNAseq and MS-proteomics. However, their has been previous support for the “single isoform” idea on the basis of transcript data alone . An ancillary conclusion from this paper, generally overlooked in terms of its significance, was that when CDS length was taken into account approximately 50% of major transcripts did not corresponding to the ‘canonical’, max-exon, transcript as annotated in Swiss-Prot. This crucial topic is further discussed in .
 Bonner, T.I. (2014). Should pharmacologists care about alternative splicing? IUPHAR Review 4. Br J Pharmacol. Mar;171(5):1231-40. doi: 10.1111/bph.12526. PMID: 24670145.
 Tress et al. (2016). Alternative Splicing May Not Be the Key to Proteome Complexity. Trends Biochem Sci. Sep 16. doi: 10.1016/j.tibs.2016.08.008. PMID: 27712956.
 Gonzàlez-Porta et al. (2013). Transcriptome analysis of human tissues and cell lines reveals one dominant transcript per gene. Genome Biol. Jul 1;14(7):R70. doi: 10.1186/gb-2013-14-7-r70. PMID: 23815980.
 Will the real cannoical protein please stand up.
Comments by Chris Southan