IRC24 (P40580) vs Nre1 (P40579): Paralog Pocket Specificity — Curator Report OpenScientist openscientist-autonomous 7 artifacts 2026-10-09T18:58:29.261669

OpenScientist AF P40579 F1 confidence v6 (application/json)

OpenScientist AF P40580 F1 confidence v6 (application/json)

OpenScientist AF P40580 F1 predicted aligned error v6 (application/json)

OpenScientist final report (text/html)

OpenScientist final report (application/pdf)

OpenScientist go decision and evidence matrix (text/csv)

OpenScientist irc24 nre1 structural analysis (text/markdown)

IRC24 (P40580) vs Nre1 (P40579): Paralog Pocket Specificity — Curator Report

Gene: IRC24 / YIR036C, Saccharomyces cerevisiae S288C (NCBITaxon:559292); UniProt P40580 ("Benzil reductase ((S)-benzoin forming) IRC24", BZRD_YEAST).
Paralog: Nre1 / YIR035C; UniProt P40579 (uncharacterized oxidoreductase).
Hypothesis slug: paralog-pocket-specificity-alphafold.
Focus type: free_text.

Executive Judgment

Verdict: Partially supported — with the two halves of the seed hypothesis landing differently.

"The paralogs have diverged in substrate preference" → REFUTED by the structural evidence. The AlphaFold Irc24 model and the experimental Nre1 crystal structure superpose at 1.36 Å CA RMSD, and the substrate pocket is more conserved than the protein overall (15/18 residues identical, 83%) with a strictly conserved catalytic Ser–Tyr–Lys triad. This "essentially shared pocket" is exactly the outcome the seed flagged as plausible given the near-identical benzil kinetics, and it is what the data show. The only hint of divergence is subtle: Irc24's pocket is slightly more hydrophobic (loss of the Nre1 Asp138 carboxylate; Irc24 65% vs Nre1 50% hydrophobic lining), a second-shell change unlikely to constitute a specificity switch.
"Comparing the structures identifies the natural substrate class" → PARTIALLY RESOLVED (narrowed, not named). Structure + Foldseek constrain the chemistry but cannot name a single physiological substrate. A genuine Foldseek run shows every nearest neighbour with a known substrate reduces a ring/cyclic carbonyl (diaryl diketone, bicyclic alkaloid, pterin, alicyclic ketone, steroid); no aliphatic-dicarbonyl (methylglyoxal-type) reductase appears. Of the four options the seed asked us to rank, methylglyoxal is the least supported, and a flat/fused aromatic ring carbonyl (e.g., a quinone such as 1,4-naphthoquinone) or another ring carbonyl is the best geometric fit for the large, predominantly hydrophobic, aromatic-lined pocket. This is inference from pocket properties + neighbour annotations, not a docking calculation; no substrate-bound structure of either paralog exists.

Most important caveats. (a) There is no experimental structure of Irc24 itself — all Irc24 geometry is AlphaFold v6 (mean pLDDT 96.4; pocket pLDDT 93.6–98.8, so confidence is not the limiting factor). (b) Neither Nre1 structure contains a substrate (3KZV is apo + glycerol; 6UHX holds only NADP), so the pocket is inferred from the cofactor site and fold. (c) The closest natural-substrate homologs (sepiapterin reductase, tropinone reductase) act on substrates yeast does not make, so homology transfer of a specific substrate is unsafe. An honest answer is: the structure cannot upgrade the annotation from "carbonyl reductase (NADPH)" to a specific physiological substrate.

Evidence Matrix

#	Citation	Evidence type	Supports/Refutes/Qualifies	Claim tested	Key finding	Context	Confidence & limitations
1	In-session computation (numpy NW-BLOSUM62 + Kabsch); AlphaFold DB AF-P40580-F1 v6; RCSB 6UHX	structural/computational	Refutes "diverged pocket"	Do Irc24 and Nre1 pockets differ?	53.4% global identity; 1.36 Å CA RMSD; pocket 15/18 identical; Ser–Tyr–Lys triad conserved	Yeast SDR; AF model vs 2.75 Å crystal	High for "shared pocket"; AF model (no Irc24 crystal); no bound substrate
2	In-session Foldseek web server (3diaa; afdb-swissprot/afdb50/pdb100)	structural/evolutionary	Qualifies substrate class	Which substrate class fits?	Neighbours = benzil reductase (6YC8 KRED1-Pglu), tropinone reductase, sepiapterin reductase, 17β-HSD14, cyclo(pent/hex)anol DH; no methylglyoxal reductase	Query = Irc24 AF v6	High that neighbours are ring-carbonyl SDRs; annotation-transfer ≠ proof of yeast substrate
3	PMID 11796169 (Maruyama 2002)	direct assay	Supports core MF	Is YIR036C a benzil reductase?	"yeast YIR036C protein … also reduced benzil to (S)-benzoin in vitro"; SDR related to sepiapterin reductases	S. cerevisiae ORF, recombinant, NADPH	Direct in-vitro; benzil is synthetic/non-physiological
4	PMID 33486371 (Rabuffetti 2021)	direct assay + structure	Qualifies substrate class	What do benzil reductases prefer?	"Benzil reductases are dehydrogenases preferentially active on aromatic 1,2-diketones"; crystal 6YC8, prefers bulky aromatic substrates	Pichia glucozyma KRED1-Pglu, 1.77 Å	Closest characterized functional analog; different organism
5	PMID 26377422 / 26952764 (Contente 2015/2016)	direct assay	Qualifies	Pocket accommodates bulky aromatics?	KRED1-Pglu "prefers space-demanding substrates"; orientation in binding pocket sets stereochemistry	Pichia benzil reductase	Supports large hydrophobic aromatic pocket
6	PMID 11745140 (Maruyama 2001)	sequence/evolutionary	Supports SDR family	Family placement	Benzil reductase is "a novel short-chain dehydrogenases/reductase"; homology to YIR036C and sepiapterin reductases	B. cereus gene	Orientation-level family evidence
7	RCSB 3KZV (apo) & 6UHX (NADP) headers; in-session HETATM scan	structural	Qualifies (negative)	Is a substrate crystallized?	3KZV: only glycerol+water; 6UHX: only NADP — no substrate in any structure	Nre1 crystals, 2.0 / 2.75 Å	Hard limit on naming a substrate from structure

GO Curation Implications (leads — require curator verification)

Molecular Function (MF). Evidence supports a retained but generalized MF. The obsoleted benzil-specific term was correctly removed (benzil is synthetic). The defensible computed/experimental MF is a carbonyl/ketone reductase, NADPH-dependent activity:
- Candidate MF: GO:0004090 carbonyl reductase (NADPH) activity (or the NADP-oxidoreductase parent GO:0016616 "oxidoreductase activity, acting on the CH-OH group of donors, NAD(P) as acceptor"), with evidence code IDA (PMID 11796169, in-vitro benzil reduction) for the activity and ISS/ISA for the structural class. The structure does not justify a more specific substrate term (e.g., a quinone- or steroid-specific MF) — do not add one as asserted.
- Avoid "protein binding" — a specific oxidoreductase MF is supported.

Biological Process (BP). No BP is supported by direct evidence. The "IRC24 = increased recombination centers" gene name derives from a loss-of-function screen phenotype, a downstream/pleiotropic readout, not the molecular function; do not annotate a DNA-repair/recombination BP from the enzymatic data. Leave BP unknown / non-core pending a physiological substrate.

Cellular Component (CC). Cytoplasm (GO:0005737) is consistent with both paralogs being soluble cytoplasmic SDRs (3KZV titled a "cytoplasmic protein"); retain if already supported by localization evidence.

Paralog note for the curator: because the pocket is ~83% identical and the fold is near-identical, any substrate/function annotation transferred to one paralog should be applied symmetrically or flagged, and benzil-reductase annotations on either gene should carry a "non-physiological substrate" qualifier.

Mechanistic Scope

Direct molecular activity (what the structure tests): NADPH-dependent reduction of a carbonyl (ketone/diketone) at the SDR catalytic centre (Tyr157/Ser143/Lys161 in Irc24), with the substrate carbonyl positioned over the nicotinamide C4 of NADP. This is the gene product's primary biochemical function.
Not direct / out of scope: the "increased recombination centers" phenotype (loss-of-function, pleiotropic), any genome-stability role, and any specific physiological pathway — none are established and none follow from the enzymology.

Conflicts and Alternatives

Against "diverged specificity": identical fold, 83%-identical pocket, conserved catalytic triad, indistinguishable cavity volume, and the reported near-identical benzil kinetics all point to a shared pocket. The seed's primary framing (divergence) is contradicted.
Paralog/annotation carry-over risk: P40580 is annotated "benzil reductase" in UniProt while P40579 is "uncharacterized"; given 52% identity and a shared pocket, this asymmetry is likely annotation history, not biology. Curators should treat benzil activity as a shared, in-vitro, non-physiological property.
Homology trap: nearest natural-substrate neighbours (sepiapterin reductase → tetrahydrobiopterin pathway; tropinone reductase → plant alkaloids; 17β-HSD14 → steroids) have no substrate counterpart in S. cerevisiae, so their annotations cannot be transferred as the yeast substrate.
Alternative substrate hypotheses not excluded: a quinone (e.g., 1,4-naphthoquinone; fits a flat hydrophobic aromatic pocket and is chemically plausible for oxidative-stress/quinone handling) is the best geometric match among the four options, but this is not proven; a broader "ring carbonyl" class remains open.

Knowledge Gaps

No physiological substrate identified. Checked: structure (no bound substrate in 3KZV/6UHX), Foldseek neighbours (all exogenous or non-yeast substrates), literature (only benzil assayed for the yeast protein). Matters because GO MF specificity and any BP depend on it. Resolve with substrate screening (see below).
No Irc24 experimental structure. Checked: AlphaFold v6 only (high confidence, pLDDT ~96). Matters marginally — confidence is high — but a crystal or a substrate co-complex would settle pocket geometry and the Asp138→Gly145 effect.
Functional consequence of the Nre1 Asp138 / Irc24 Gly145 difference. Checked: it is the single non-conservative near-pocket change (4.8 Å, second shell). Matters only if the paralogs are subtly specialized. Resolve by reciprocal point-mutant kinetics.
Quantitative pocket volume/druggability. Checked only with a crude grid (relative comparison valid, absolute unreliable); fpocket/CASTp/SiteMap were not run. Resolve with a dedicated cavity tool for a defensible volume and a docking panel (naphthoquinone vs methylglyoxal vs benzil vs a steroid).

Discriminating Tests

Substrate panel kinetics on recombinant Irc24 and Nre1 in parallel: 1,4-naphthoquinone and other quinones (menadione, juglone), methylglyoxal, a model steroid/ring ketone, vs benzil — measure k_cat/K_M and NADPH vs NADH. This directly tests "shared vs diverged" and names the preferred class.
Co-crystal or cryo-soak of Nre1 (or Irc24) with a candidate substrate + NADP(H) to define the real pocket.
Reciprocal mutants Irc24 G145D and Nre1 D138G; assay for any substrate-preference shift.
Dedicated pocket/docking analysis (fpocket/CASTp + AutoDock Vina) on the AlphaFold Irc24 model with the four candidate ligands for a computed ranking to replace the current inference.
Genetics/metabolomics: irc24Δ nre1Δ (single and double) metabolite profiling and quinone/oxidative-stress sensitivity, to connect the enzyme to a pathway.

Curation Leads (require curator verification)

Action on the review: Treat the seed's "diverged substrate preference" claim as not supported; record the computed "shared pocket" result instead. Treat "identifies the natural substrate class" as narrowed, not resolved (ring/fused-ring carbonyl favoured; methylglyoxal disfavoured).
Candidate MF term: GO:0004090 (carbonyl reductase (NADPH) activity) or the generalized NAD(P) CH-OH oxidoreductase parent — IDA from PMID 11796169; keep a "benzil = non-physiological substrate" note. Do not assert a quinone/steroid-specific MF.
BP/CC: BP unknown (do not infer recombination/DNA-repair from the gene name); CC cytoplasm if localization-supported.
Candidate references + exact snippets to verify:
PMID 11796169 — "We isolated the genes encoding yeast YIR036C protein and gerbil sepiapterin reductase, and both recombinant proteins also reduced benzil to (S)-benzoin in vitro."
PMID 33486371 — "Benzil reductases are dehydrogenases preferentially active on aromatic 1,2-diketones."
Suggested curator questions: Is there localization evidence for P40580/P40579? Is the benzil annotation on P40580 experimental (IDA) or transferred? Should P40579 receive the same generalized carbonyl-reductase MF given the shared pocket?
Suggested experiments: the substrate panel and reciprocal-mutant kinetics above.

Tools and database versions actually invoked (with honest limitations)

AlphaFold DB REST API + file server, model AF-P40580-F1 v6 (modelCreatedDate 2025-08-01; AlphaFold Monomer v2.0 pipeline; globalMetric/mean pLDDT 96.38) and AF-P40579-F1 v6 (96.62). Note: the seed referenced v4; the current release is v6.
RCSB PDB coordinate files 3KZV (2.00 Å, apo) and 6UHX (2.75 Å, +NADP).
In-session Python (numpy) for parsing, Needleman–Wunsch (BLOSUM62), Kabsch superposition, pocket detection, crude cavity grid, polarity. Biopython and Phenix were not applied to the Irc24 model: the code-executor sandbox is ephemeral and cannot share files with the Phenix-tool environment, and Biopython is not an allowed import there. (RCSB PDBs are staged in the job data/ dir if a curator wishes to run phenix.molprobity/phenix.superpose_pdbs on the experimental structures.) The absolute cavity volume is an overestimate from a permissive enclosure criterion and is reported only as a relative paralog comparison.
Foldseek web server (search.foldseek.com API, mode 3diaa; databases afdb-swissprot, afdb50, pdb100) — real output reproduced in provenance/irc24_nre1_structural_analysis.md.
PubMed via search_pubmed.

Provenance (computed metrics, full per-residue pocket table with pLDDT, and the Foldseek hit list) is saved to provenance/irc24_nre1_structural_analysis.md.