Consistency: PARTIAL CONTRADICTION (same axis as VPS37B/C), with the weakest gene-specific evidence of the family. ESCRT-I membership consistent: review ACCEPTs GO:0000813, but notes VPS37D family assignment is largely by sequence similarity (PMID:15218037 studied VPS37B; only UniProt/ComplexPortal + PMID:18005716/22405001 support membership). Autophagy diverges: review MARK_AS_OVER_ANNOTATED GO:0016236 and GO:0090148 (membrane fission), declining GO:0000045. PN still projects GO:0000045 to VPS37D.
PN story / NEW pressure: Over-reaches for autophagy. PN's group GO:0000045 asserts an autophagosome-assembly role with no VPS37D-specific support at all (no phagophore-closure, membrane-scission, or even direct MVB assay for this paralog). ESCRT-I already captured; review even cautions against overstating VPS37D experimental testing. No NEW term beyond GO:0000813.
Mapping strategy: Flag (strongest case in family). Leaf→GO:0000813 acceptable (in GOA, defensible by homology). Group→GO:0000045 clearly over-reaches — VPS37D is the least-characterized paralog; this is exactly the shared-node over-propagation the TOMM20/HSPA8/RAB7A precedent rejects. GO:0000045 must not propagate to VPS37D.
Evidence alignment: Partial. PN cites only PMID:32424346 (VPS37B headpiece), which the review explicitly says should not transfer membrane-fission/autophagy to VPS37D. Review adds PMID:22405001 (UBAP1 coassembly) not in PN.
Verdict: Consistent on ESCRT-I; GO:0000045 projection over-reaches most strongly here. Recommended edits: none to YAML (review correctly omits GO:0000045); flag PN sealing-group→GO:0000045 as non-propagating to VPS37D.