Consistency: Mismatch in emphasis. Deep-research notes and review converge on a nuclear J-domain+RRM protein whose best-supported (convergent IBA + EFTUD2 IPI + nuclear-speckle) function is spliceosome association / spliceosomal complex disassembly (yeast Cwc23 ortholog) plus RNA binding. The PN node frames it solely as an HSP70 co-chaperone. The HSP70-binding role is plausible from the J domain but is NOT demonstrated and NOT the review's core; the review never asserts Hsp70 binding. So PN and review diverge on what the gene principally does.
PN story / NEW pressure: PN proposes GO:0030544 (verified real). There is no existing Hsp70-binding annotation and no direct evidence one exists for DNAJC17; the "more_specific_than_existing_goa" tag is questionable (no existing parent MF in GOA to be more specific than — existing MF terms are RNA binding/nucleic acid binding, a different axis). Family-level inference only; over-reaches as a confident projection.
Mapping strategy: Node heuristic is fine generically, but for DNAJC17 the projection is purely domain-based and the gene's real placement is in the spliceosome (Translation/RNA branch), not captured here. PN-projected Hsp70 binding is orthogonal to (not broader/narrower than) the review's RNA-binding/spliceosome core. Treat as family inference.
Evidence alignment: Divergent. Review cites EFTUD2 interactome papers (PMID:35271311 OpenCell, PMID:40205054 cell maps) and the IBA spliceosome-disassembly inference; PN node carries no DNAJC17-specific reference, only the family heuristic. No shared functional paper supports Hsp70 binding.
Verdict: Plausible family inference but over-reaches as confident projection; gene's real function (spliceosome/RNA) sits outside the HSP70-cochaperone node. Recommended edits: [MAP] mark DNAJC17's GO:0030544 as family/domain-level inference (no experimental HSP70 binding); correct the "more_specific_than_existing_goa" tag (existing GOA MF is RNA binding, not an HSP70 parent).