Re-annotate um_tcga_gdc with vibe-vep

[inodb]

Re-annotates Uveal Melanoma (TCGA GDC) mutations with vibe-vep.

Uses GENCODE v46 transcript annotations instead of genome-nexus/mskcc isoforms.

[inodb]

vibe-vep re-annotation: protein change differences

78 out of 1490 variants have different HGVSp_Short values compared to the original genome-nexus/mskcc isoform annotations. Categories:

Gene reassignment (~10 variants)

vibe-vep picks a different canonical transcript, mapping to a different gene:

• MIR3918 → SYTL3, MIR6720 → FOXF2, RHBDF1 → SNRNP25, ZSWIM3 → ZSWIM1, ANAPC11 → ALYREF, SUPT5H → TIMM50, PARD6A → ACD

Position shifts (~20 variants)

Same gene/transcript, different residue numbering due to updated transcript versions:

• PLCB4 p.D630N → p.D612N, CFAP70 p.A654T → p.A584T, MUCL3 p.S377C → p.S1253C, ARID1B p.P631Q → p.P701Q

Splice → synonymous reclassification (~8 variants)

Old X_splice calls become synonymous with updated annotation:

• SPAG5 p.X580_splice → p.A580=, ZMAT1 p.X41_splice → p.D41=, DDX51 p.X558_splice → p.L558=

Frameshift refinements (~15 variants)

Slightly different frameshift descriptions due to updated transcript versions:

• BAP1 p.S485Kfs*82 → p.S489Rfs*78, BAP1 p.D672Afs*17 → p.C676Sfs*13
• GPC5 p.S563Kfs*22 → p.I562Yfs*23, SYNE2 p.I6563Hfs*9 → p.A6562Rfs*10

Non-coding stop-gains cleared (~15 variants)

p.*N* annotations on pseudogenes/ncRNAs correctly become empty (non-coding transcripts shouldn't have protein annotations):

• XIST, MIR125A, MIR224, MIR518E, SNORD116-11, NACA, CASTOR3P, etc.

Clinically notable

• EIF1AX p.K7_G8delinsR → p.G9del — key uveal melanoma driver
• BAP1 frameshift changes — key uveal melanoma tumor suppressor