Lightnews — Scholar-powered news

Michael Tress @michaeltress.bsky.social · 6d

... selection in mammals. I guess any of the three is OK, but if I had to choose (and APPRIS has to choose), I would be inclined to choose ENST00000355480 over the 1319 aa version, given the detailed RNAseq evidence. The longest isoform is alternative for me. Another one to change for APPRIS.

Michael Tress @michaeltress.bsky.social · 6d

... evidence for the other two, the longest transcript almost exclusively in spinal cord, ENST00000355480 in liver and artery where it is supposed to be most expressed. Not much proteomics support here either, though there are some non-tryptic peptides. All three isoforms are under purifying ...

1

Michael Tress @michaeltress.bsky.social · 6d

Also, I had forgotten about this one. These two isoforms differ at the N-termanal and there's a third (from ENST00000355480) that has 1519 residues. RNAseq strongly supports the shorter N-terminal (1339aa), but then it would. I can't see proteomics evidence for this isoform. There is some RNAseq ...

1

Michael Tress @michaeltress.bsky.social · 6d

Thanks for all the work you put into this, it has been very useful. We think it has confirmed that @appris.bsky.social principal selection is mostly working as it should, and it has helped find problematic genes. We have changed 5 principal isoforms and the labelling of alt. isoforms in 6 more.

Michael Tress @michaeltress.bsky.social · 8d

MIA2 is a struggle because there isn't much data for any of the annotated N-terminals. The RNAseq support for the shorter transcripts is not totally believable, especially since they have practically no peptide support.

Michael Tress @michaeltress.bsky.social · 8d

There are times when you can only use one isoform/transcript per gene, so some rules have to be applied.

But there are certainly plenty of genes that clearly have multiple important isoforms, which is why APPRIS has TRIFID scores. Even then you get isoforms like the one in RPGR that buck the trend

1

Michael Tress @michaeltress.bsky.social · 8d

APPRIS went with the 79aa isoform that has RNAseq as well as proteomics support.

Michael Tress @michaeltress.bsky.social · 8d

We had already changed the APPRIS principal manually to the brain specific AS variant. We think the MANE Select variant should have a different N-terminal.

Michael Tress @michaeltress.bsky.social · 8d

MANE and APPRIS both support 235 aa, so all the major annotations in agreement.

Michael Tress @michaeltress.bsky.social · 8d

Haha. It is a bit messy, but I can assure you I have seen MUCH worse. Superficially it appears to have been two genes bolted together, but it appears to exist across mammals at least. RNAseq supports 792aa and 804aa isoforms, but it would, they are shorter. I think probably all three are used.

1

Michael Tress @michaeltress.bsky.social · 8d

Agree that this looks likely. But the first ATG is conserved all the way across mammals, so I can't see this being changed any time soon. Especially since GENCODE no longer have anyone to point this out.

The 2nd ATG is conserved too and its Kozak sequence is largely untouched, unlike the first ATG

Michael Tress @michaeltress.bsky.social · 8d

There is also a 815aa isoform that fits the proteomics and RNAseq data perfectly. It skips the retained introns that produce the 1020 aa and 1152aa isoforms. The 1152aa version has no evidence, but has oodles of clinical variants and is expressed in retinal cilia pmc.ncbi.nlm.nih.gov/articles/PMC...

Impaired glutamylation of RPGRORF15 underlies the cone-dominated phenotype associated with truncating distal ORF15 variants

Retinitis pigmentosa GTPase regulator (RPGR) disease variants lead to considerable phenotypic heterogeneity in terms of relative involvement of rod and cone photoreceptors. We have identified a clear ...

pmc.ncbi.nlm.nih.gov

Michael Tress @michaeltress.bsky.social · 8d

The APPRIS principal has the extra exon. We will change it to agree with the MS transcript

Michael Tress @michaeltress.bsky.social · 8d

APPRIS and MANE now support a 339aa principal isoform. It has an extra exon vs. the transcript that produces 321 aa isoform. RNAseq suggests that both 339 and 321aa isoforms are equally valid, the exon is largely skipped in brain, and not elsewhere.

Michael Tress @michaeltress.bsky.social · 8d

Both APPRIS and MANE have 801 aa isoform as principal, so all in agreement here now.

Michael Tress @michaeltress.bsky.social · 8d

Yeah, no support at all for the MANE Select. RNAseq and conservation support the 73aa UniProt/APPRIS isoform too.

Michael Tress @michaeltress.bsky.social · 8d

RNAseq is in agreement. APPRIS had already switched its principal to the 431aa isoform, MANE still chooses the alternative which has some (low) expression in brain.

Michael Tress @michaeltress.bsky.social · 8d

APPRIS/MANE have both already switched to the correct 525aa isoform

Michael Tress @michaeltress.bsky.social · 8d

The upstream ATG is protein coding conserved across primates, but is only expressed in testis (and no peptides as noted), so the 287aa isoform is clearly principal. APPRIS changed.

Michael Tress @michaeltress.bsky.social · 8d

APPRIS had already switched to the 374aa isoform, but the RNAseq data does not agree with the proteomics data. It has inclusion and skip at about 50-50.

Michael Tress @michaeltress.bsky.social · 8d

I like this one! You are absolutely right that there is more evidence for the 805aa isoform. But there are two NAGNAG splice events in RASAL1 and the extra amino acid exon has way more support in splice events. Which makes the 806aa isoform the principal. APPRIS will change to reflect this.

Michael Tress @michaeltress.bsky.social · 8d

However, this paper shows strong evidence for three N-terminal AS variants in KCNIP1 in different regions in the brain, the 227aa isoform (UniProt), 216aa (MANE) and 225aa (APPRIS principal), so I am inclined to leave things as they are.

Michael Tress @michaeltress.bsky.social · 8d

I am not sure what to make of this one. There are a lot of conserved N-terminal exons. It is brain expressed. I wouldnt put much stock in the proteomics data (non-existent for any isoform) nor Alphafold (no structure for any). The RNAseq evidence strongly supports the 227aa variant. 1/2

1

Michael Tress @michaeltress.bsky.social · 8d

It does have RNAseq support though (inclusion is about 50/50).

Michael Tress @michaeltress.bsky.social · 8d

LCOR is one of only about 20 annotated splice variants in which one set of Pfam domains is swapped for another set (cf CUX1, DST, etc). Different main isoforms is OK.