Hypothesis explorer

Evidence quoted from the run

The H1 decision rule required ≥2 LDAM markers upregulated and ≥2 DAM2 markers downregulated at FDR<0.05; neither criterion was met. H1 is therefore not supported.

Adjudication note

0/7 markers FDR<0.05; DAM2 trended UP. Tested only as binary APOE4 carrier (n=25) vs non-carrier (n=59) because CellxGene SEA-AD lacks e4/4 vs e3/3 genotype; bootstrap log2FC instead of DESeq2; Morabito replication impossible (no APOE column). Strict reading could call this NOT_TESTED_DATA (diluted proxy contrast). Knowledge-graph status = refuted.

Evidence quoted from the run

The partial regression coefficient for Braak stage was β_tau=+0.031 (95% bootstrap CI: −0.007 to +0.069; p=0.143; Fig. 4c)

Adjudication note

Report: 'H2 is therefore not supported.' Direction positive; plan rated H2 adequately powered (MDE r~0.29) but report says post hoc n~290 needed for observed r=0.162; age covariate dropped (categorical bins). Knowledge-graph last status = inconclusive.

Evidence quoted from the run

Because the CellxGene object provides only binary carrier status, the test reduces to a two-group comparison equivalent to Mann-Whitney U.

Adjudication note

Three-genotype dose-response untestable (binary APOE4 only). Proxy 2-group, Braak-UNadjusted test: LDAM p=0.006 (Bonf. 0.012), DAM2 p=0.887. Report relabels this as 'exploratory, not pre-registered' although H3 is in the PLAN-stage hypothesis list. Knowledge-graph status = supported (discrepancy). If proxy counted as a test: at most PARTIALLY_SUPPORTED.

Evidence quoted from the run

upregulates APOE transcription.

Adjudication note

Raised in EXECUTE stage 2 (marked cheap to test); no later execute stage; data available but never run.

Evidence quoted from the run

Accessing fine-grained ε4/4 vs ε3/3 homozygous genotype from the SEA-AD Synapse release (syn64410371) and re-running the H1 pseudobulk DE contrast

Adjudication note

Listed only as a next step; fine-grained genotype absent from the fetched CellxGene object and the Synapse metadata was never fetched.

Evidence quoted from the run

with the same 84 donors.

Adjudication note

Never run; data available.

Evidence quoted from the run

Under that denominator, the observed proportion is 13/75=17.3%, and the binomial test against p₀=0.25 gives p≈0.87 (FAIL).

Adjudication note

Criterion 1 fails under pre-registered n=75 denominator (only 33/75 loci had any peak); among 33 testable loci 13 MG-dominant, binomial p=0.048 (post hoc conditioning). Criterion 2 (SPI1) untestable: JASPAR/FIMO unavailable, logistic proxy circular/non-convergent. OmiCairn stage 4 analysis labelled 'PARTIAL' and knowledge-graph status = supported (discrepancy; code review & report call it FAIL).

Evidence quoted from the run

GWAS-overlapping peak accessibility was not stage-dependent (early vs late Δ=−0.008, 95% CI −0.023 to 0.007, p=0.881), refuting H2.

Adjudication note

Late-stage numerically higher (opposite). DAM-marker Stage2>Stage1 contrast 'significant' but flagged by code review as bootstrap collapse at 3-5 donors/group. Only 20 Morabito donors (plan: underpowered, d>=1.45); SEA-AD fallback not run (Synapse auth). Knowledge-graph nodes: one inconclusive, one refuted.

Evidence quoted from the run

Number of GWAS loci where each cell type is dominant (MG: 13, ASC: 7, EX: 3, ODC: 2; 8 below accessibility floor).

Adjudication note

UNCERTAIN: no pre-registered threshold and the planned chi-square/pairwise Fisher tests were never run; verdict is descriptive only (observed order MG>ASC>EX>ODC, EX not second). Defensible alternative: INCONCLUSIVE.

Evidence quoted from the run

This CI excludes 0, supporting directional microglial enrichment among loci with detectable accessibility.

Adjudication note

Delta=0.273, 95% CI [0.071, 0.485], p=0.004, but only among 33/75 testable loci; report notes compositional non-independence and that the stricter permutation-null comparison is not significant (Delta=0.029, p=0.483).

Evidence quoted from the run

The SEA-AD atlas [9]—84 donors spanning the full severity spectrum with paired snATAC-seq and quantitative neuropathology—is the appropriate resource to resolve H2 definitively.

Adjudication note

Raised twice (stages 2 and 4). Morabito early-vs-late CI includes 0; SEA-AD snATAC blocked by Synapse authentication.

Evidence quoted from the run

**Use fine-mapped credible sets** from Bellenguez 2022 rather than lead SNPs.

Adjudication note

Run halted (orphaned) at claim-review stage 7 before any follow-up execute; credible sets are public.

Evidence quoted from the run

H1 therefore fails on all three pre-registered sub-criteria.

Adjudication note

Final EXECUTE stage 9: C3-ITGAM rank 257 (p=0.565), SPP1-ITGA9 rank 45 and DOWN, CXCL10-CXCR3 absent from CellChatDB. Braak-monotone component untestable (no Braak in GSE157827; binary AD-vs-NC only). Global donor-label permutation p=0.494; report calls it 'an underpowered null result, not a powered negative' -> some may prefer INCONCLUSIVE. Graph H1 nodes last status inconclusive; stage 9 graph update = refutes.

Evidence quoted from the run

H2 as pre-registered (direct overlap) is therefore untestable with this resource

Adjudication note

OmiCairn/report label it NOT SUPPORTED (OR=0.0, CI [0, 3.59], p=1.0) but only 2 GWAS genes are in the 305-gene CellChatDB background, i.e. a structurally degenerate test; Braak III-IV stratum also unavailable. Alternative: INCONCLUSIVE. Graph = refuted.

Evidence quoted from the run

could not be tested in this study because GSE157827 covers only PFC and the SEA-AD multi-region atlas (Synapse syn64410371; [19]) was not fetched for this analysis

Adjudication note

Single-region atlas; multi-region data never fetched.

Evidence quoted from the run

this should be tested in SEA-AD where continuous neuropathology scores are available

Adjudication note

Needs Braak-resolved data; SEA-AD never fetched.

Evidence quoted from the run

in a multi-donor dataset with both BBB markers and Braak staging

Adjudication note

No Braak-staged dataset used; JAM2->ITGB1 only nominal p=0.021 uncorrected in GSE157827.

Evidence quoted from the run

This reformulation of H2 is the principal actionable next step.

Adjudication note

Never run (different framework needed); data available.

Evidence quoted from the run

SEA-AD with n > 400 donors would test this definitively.

Adjudication note

Present only in an EXECUTE-stage graph update (not merged into the knowledge graph); larger cohort never fetched.

Evidence quoted from the run

MDK is a heparin-binding growth factor with neuroprotective roles

Adjudication note

Only in stage 9 graph-update record (not in knowledge graph); flagged cheap to test, never followed up.

Evidence quoted from the run

Semaphorin 5A is implicated in neural circuit refinement

Adjudication note

Only in stage 9 graph-update record (not in knowledge graph); never followed up.

Evidence quoted from the run

**H1 verdict: NOT SUPPORTED.** No gene reaches BH-FDR < 0.05.

Adjudication note

Latest (008, GSE147528 primary, 2,009 RORB+ cells/10 donors): CACNA1C/D opposite sign, CIs include 0; PVALB log2FC -1.18 CI excludes 0 but FDR 0.865 (t-test NaN issue). Report calls it a powered negative. Atlas roles swapped vs plan (GSE138852 had only 16 RORB+ cells).

Evidence quoted from the run

failing the pre-registered H2 criterion of a positive slope at p < 0.05

Adjudication note

Slope -0.006/Braak (CI [-0.018, +0.002], p=0.213), opposite direction; n=10 donors, Braak 0/2/6 only (plan: borderline power).

Evidence quoted from the run

An exploratory excitability gene module score is positive in direction (d = +0.269) but spans zero at the donor level of inference (approximate CI [−0.362, +0.900])

Adjudication note

Excitability had the largest |d| but below the 0.3 SD benchmark; cell-level bootstrap CIs anti-conservative, donor-level CI spans 0. Knowledge-graph node ('excitability module higher') = supported (discrepancy).

Evidence quoted from the run

This is a **Tier-A negative finding**: the primary atlas does not contain enough RORB+ neurons for valid pseudobulk DE

Adjudication note

GSE138852: 16 RORB+ nuclei, 1/6 pools >=5 cells (pre-registered Tier-A negative). Leiden resolution sweep never run (OMP crash); threshold-based annotation used. Knowledge-graph status = refuted.

Evidence quoted from the run

The ISR module score was lower in RORB+ than RORB- neurons (d = −0.056, cell-level CI [−0.100, −0.008]; approximate donor-level CI [−0.678, +0.566]; Fig. 5a)

Adjudication note

Knowledge-graph status = supported (cell-level CI); donor-level CI spans 0. 'At all Braak stages' not tested stage-stratified.

Evidence quoted from the run

CACNA1C and CACNA1D show small NEGATIVE log2FC (opposite the H1 prediction of upregulation in RORB+).

Adjudication note

Acted on in EXECUTE stage 8 (atlas switch); CACNA1C/D FDR=1.00.

Evidence quoted from the run

per-gene identification of driver genes within the module has not been performed

Adjudication note

Data available; not run.

Evidence quoted from the run

For the pseudobulk DE, use pydeseq2 (once OMP issues are resolved) which handles sparse genes more robustly than the simple t-test.

Adjudication note

Technical: pydeseq2 not run because of native OMP/BLAS library crashes on the analysis platform.

Evidence quoted from the run

H1 is INCONCLUSIVE because the Kamath 2022 SN snRNA-seq (GSE178265, approximately 400,000 nuclei) requires dbGaP controlled-access approval

Adjudication note

dbGaP controlled access; only descriptive surrogate (GSE184950, 4 samples), AUROC never computed.

Evidence quoted from the run

The primary analysis could not be executed as pre-registered because none of the five TFEB target genes appear in the NanoString NS_HS_NEUROPATH_V1.0 panel of GSE167335

Adjudication note

Report/analysis label 'NOT SUPPORTED' and graph = refuted, but this rests on a post-hoc V-ATPase surrogate (8/9 up, donor-level permutation p=0.333, N=2 donors). GBA1-N370S arm blocked (no isogenic dataset); synergy arm untestable.

Evidence quoted from the run

H3 is BLOCKED: no matched iPSC-DA RNA-seq + lysosomal proteomics dataset with simultaneous measurement of both modalities was accessible.

Adjudication note

Reversal arm also untested (no treatment data).

Evidence quoted from the run

this design is fundamentally underpowered for significance testing.

Adjudication note

8/9 genes up, mean log2FC +0.485 (CI -0.454 to 1.471), exact donor permutation p=0.333 = minimum achievable at N=2+2. The knowledge-graph node marked supported came from the pseudo-replicated stage 3 analysis; the latest stage 8 node = inconclusive. MLi-2 reversal part never testable (no data).

Evidence quoted from the run

The GBA1 N370S arm of H2 could not be tested: no GBA1 N370S isogenic iPSC-DA RNA-seq dataset was found in accessible GEO deposits

Adjudication note

No adjudication note was recorded.

Evidence quoted from the run

N ≥ 5 independent biological replicates at the donor level is necessary

Adjudication note

Methodological/future-data statement; no suitable dataset fetched.

Evidence quoted from the run

Per the pre-registered decision rule, this outcome is classified as **partial sharing (DLB+PDD)**

Adjudication note

DLB g=1.55 & PDD g=2.64 (FDR 0.014); AD g=1.20 FDR 0.343 (n=4 vs 4, cannot reach FDR without perfect separation). Pre-specified DAM2 dissociation control failed (DAM2 co-elevated), weakening LDAM specificity. Graph per-arm nodes: DLB/PDD supported, AD inconclusive.

Evidence quoted from the run

The H2a result, as reported, is best understood as a threshold-sensitivity restatement of H1 rather than independent corroboration.

Adjudication note

OmiCairn verdict 'corroborates' (rule formally met) but H2a was a semi-circular threshold proxy (no Leiden/scCODA; 29-gene extraction) and AUCell reached FDR<0.05 for no disease, AD sign discordant. Alternative: INCONCLUSIVE.

Evidence quoted from the run

Per the pre-registration, a CI spanning 0.5 constitutes a null outcome

Adjudication note

AUC=0.60, CI [0.45, 0.76], p=0.093, n=80 (68 high / 12 low pathology); report notes class imbalance means 'not powered evidence of absence'. Knowledge-graph status = inconclusive.

Evidence quoted from the run

The APOE4-LDAM association is not replicated in this independent analysis (Fig. 4b–c).

Adjudication note

GSE303823 arm UNTESTABLE (no APOE metadata) -> that arm is NOT_TESTED_DATA; verdict rests on the pre-registered SEA-AD MTG arm: MWU p=0.150, r=-0.20 (opposite). Note SEA-AD MTG is the same object as the prior finding, not an independent cohort.

Evidence quoted from the run

H5 verdict is **inconclusive**: neither the expected AD-specific disruption nor a shared-disruption pattern was detected.

Adjudication note

All contrasts near zero (FDR=1.0); no cell-type annotation, 29-gene extraction.

Evidence quoted from the run

preserved DLB elevation (g=1.87, 95% CI [1.33, 3.79]) but rendered PDD statistically ambiguous (g=0.84, 95% CI [−0.11, 2.20])

Adjudication note

Proportion adjustment and threshold sweep robust (DLB, PDD), but PDD attenuated with 6-gene set and PMI/age/sex covariate adjustment not feasible (absent metadata).

Evidence quoted from the run

Pre-registered threshold: ρ ≥ 0.70 with CI excluding 0. Both criteria met. **H1 is supported.**

Adjudication note

rho=1.000, degenerate CI [1,1], perm p=0.0083. Caveat: literature-fallback (murine) connectivity, connectome-only model A also rho=1.0 (ceiling). Superseded sibling run parkinson-asyn-propagation-r2 with Allen connectivity found H1 NOT SUPPORTED. This run never reached write/review.

Evidence quoted from the run

Δρ(B − A) = 0.000 (95% CI [0.000, 0.000]; p_directional = 0.998)

Adjudication note

Both models at ceiling.

Evidence quoted from the run

Vocabulary mismatch (pSyn+ TrueID format not matching profile index barcodes) forced use of all 48,702 profile rows, biasing LAMDA top-3 toward abundant cell types

Adjudication note

OmiCairn: NOT SUPPORTED (overlap 0, EN ranked 3rd). But LAMDA side is an all-spots proxy, and model PFC top-3 contains DMNX_Neu/GPI_Neu (annotation artefact, judged uninterpretable in sibling run parkinson-asyn-propagation-r2). Strict reading: NOT_SUPPORTED.

Evidence quoted from the run

Maximum rank shift = 3 > pre-registered threshold of > 1.

Adjudication note

Deterministic model result under literature-fallback connectivity; reversed (0 shift) in sibling run parkinson-asyn-propagation-r2 with Allen connectivity.

Evidence quoted from the run

Models A and B tie at ρ = 1.000; Model C achieves ρ = 0.400.

Adjudication note

Descriptive; A=B at ceiling, B-C Delta 0.601 CI [0, 2] p=0.336.

Evidence quoted from the run

not meeting the pre-registered criterion that requires CI exclusion of zero

Adjudication note

Latest EXECUTE stage 19: rho=0.711 (meets point threshold), CI [-0.304, 1.000], perm p=0.094; n=5 regions, plan conceded ~50-65% power -> borderline INCONCLUSIVE.

Evidence quoted from the run

the observed Δρ is negative and the CI does not exclude zero. H2 is NOT SUPPORTED.

Adjudication note

Delta rho(B-A) = -0.207 (CI -1.304 to 0.913); connectome-only A outperforms B (0.918; 1.000 when independently optimised, post hoc).

Evidence quoted from the run

H3 is not biologically interpretable; the following analysis is reported for pre-registration completeness only.

Adjudication note

OmiCairn headline 'H3 is NOT SUPPORTED' (overlap 0, Fisher p=1.0), but report itself deems it uninterpretable (DMNX_Neu/GPI_Neu annotation artefacts in PFC; sub-SIR never reached threshold).

Evidence quoted from the run

Zero rank displacement (max |Δrank| = 0) was observed across all five regions after removing the DMNX→GPi edge (Fig. 5a).

Adjudication note

No adjudication note was recorded.

Evidence quoted from the run

The three model variants ranked as: Model A (ρ = 0.918, CI [0.745, 1.000]) > Model B (ρ = 0.711, CI [−0.304, 1.000]) > Model C (ρ undefined; Fig. 5b,c,d).

Adjudication note

UNCERTAIN mapping: H5 is descriptive ('always reportable'); verdict reflects that the primary model B did not rank best (inversion). Pairwise Delta CIs include 0; alternative INCONCLUSIVE.

Evidence quoted from the run

API Error: The socket connection was closed unexpectedly.

Adjudication note

EXECUTE stage 3 terminated on a connection error before writing results (no results/analysis records); run halted and was never re-executed.

Evidence quoted from the run

API Error: The socket connection was closed unexpectedly.

Adjudication note

EXECUTE stage 3 terminated on a connection error before writing results (no results/analysis records); run halted and was never re-executed.

Evidence quoted from the run

API Error: The socket connection was closed unexpectedly.

Adjudication note

EXECUTE stage 3 terminated on a connection error before writing results (no results/analysis records); run halted and was never re-executed. Plan already expected H3 likely INAPPLICABLE (no donor overlap between PBMC and post-mortem brain).

Evidence quoted from the run

API Error: The socket connection was closed unexpectedly.

Adjudication note

EXECUTE stage 3 terminated on a connection error before writing results (no results/analysis records); run halted and was never re-executed.

Evidence quoted from the run

API Error: The socket connection was closed unexpectedly.

Adjudication note

EXECUTE stage 3 terminated on a connection error before writing results (no results/analysis records); run halted and was never re-executed.

Evidence quoted from the run

**H1 SUPPORTED:** 9 L-R pairs are shared between PBMC CD8+T→monocyte and brain CD8+T→CNS macrophage PD-upregulated top-100 lists

Adjudication note

Rule met (9 pairs; p<1/10,000). Caveats: brain arm = Fullard non-SN fallback (no SN tissue); NATMI substituted for LIANA rank_aggregate (segmentation fault); 1,000 cells/donor subsample; hardcoded donor-disease fallback mapping.

Evidence quoted from the run

**Decision: H2 NOT SUPPORTED (p = 0.346).**

Adjudication note

Only 18 genes; run notes low power (OR=2.56, CI 0.058-18.0).

Evidence quoted from the run

**Status: INAPPLICABLE.** Per the scope stage (data profile), the parkinson-tcell-foothold cross-workspace results record contains only atlas-level summary statistics

Adjudication note

No per-donor brain vector available.

Evidence quoted from the run

"H4_reverse_permutation_p": 0.0528,

Adjudication note

DISCREPANCY: pre-registered rule (reverse count <5 AND p>=0.05) is literally met (count=4, p=0.0528), but analysis/report label it 'BORDERLINE'/'inconclusive', apparently misreading the p>=0.05 clause. Substantively weak: reverse p is near-significant. Reviewer may prefer INCONCLUSIVE.

Evidence quoted from the run

All 5 H5 "method" counts use NATMI expr_prod columns with different sorting — not true independent methods.

Adjudication note

Run/report call H5 SUPPORTED (max deviation 0), but independent methods were never run (LIANA rank_aggregate segfault) — proxy-only, technical failure.

Evidence quoted from the run

**H1 verdict: NOT SUPPORTED at the pre-registered threshold** (threshold = cosine > 0.30). The mean cosine of 0.025 does not reach 0.30.

Adjudication note

Permutation component met (p=0.001 vs null mean -0.135) but cosine CI [-0.067,0.114] includes 0; residualized delta p=0.43. CellOracle not run (velocyto build failure) - expression-stratification proxy; DMNX_Neu fallback instead of SNpc. Could arguably be PARTIALLY_SUPPORTED (one of two rule components met).

Evidence quoted from the run

**H2 NOT SUPPORTED** (ρ threshold = 0.4, p threshold < 0.05).

Adjudication note

rho=0.042; pre-flagged ~30% power.

Evidence quoted from the run

**H3 SUPPORTED** (q < 0.05). LRRK2-KO predicted targets are significantly enriched in the Hemedan 2024 Boolean-PD prodromal regulatory module (FDR q = 0.007, FE = 4.7×).

Adjudication note

Overlap is 5 genes (SNCA, LRRK2, GAK, CTSB, SCARB2) in an 11-gene HVG-restricted module; overlapping genes are themselves GWAS/KO genes (possible circularity). Proxy KO method, not CellOracle.

Evidence quoted from the run

- **SNCA-KO**: cosine = +0.235, permutation p = 0.001 (one-tailed: cosine ≥ 0.235 in null). **H4a SUPPORTED**

Adjudication note

Literal rule met. Interpretive incoherence: positive cosine means KO shifts TOWARD the PD state, yet text calls it 'protective'.

Evidence quoted from the run

- **HBB-KO**: |cosine| = 0.032 < 0.1 threshold. **H4b SUPPORTED**

Adjudication note

Report notes residualized HBB cosine rises to 0.236; raw (primary) value used.

Evidence quoted from the run

The Zenodo midbrain reference base-GRN comparison was not executed (CellOracle package unavailable due to velocyto build failure; Zenodo download was not attempted).

Adjudication note

Surrogate raw-vs-residualized comparison gave |delta|=0.118 (>0.1); report reports H5 'not formally supported' on the surrogate.

Evidence quoted from the run

adams_2024,0.9761754643911956,-0.9433853104864593,2.6413861241359333,0.8454192090937935,1.0,3,10,91,2703,False,False,-1.6

Adjudication note

Formal rule failed (0/2 atlases pass; Adams log2FC +0.98, Wang 0.0, FDR=1.0) but only 3 PD vs 10 ctrl (Adams, young+aged ctrl pooled) and 5 vs 4 (Wang) donors vs planned 14/8 and 23/9; Fullard never fetched. Severely underpowered; no analysis summary written. Strict reading would be NOT_SUPPORTED.

Evidence quoted from the run

adams_2024,tp53,-0.887683905934536,,,0.1393939393939394,0.1393939393939394,False,194

Adjudication note

Only Adams evaluable (Wang/Fullard no values); >=2-atlas rule cannot be satisfied; underpowered.

Evidence quoted from the run

ERROR loading Martirosyan count table: could not convert string to float: 'MIR1302-2HG'

Adjudication note

Data were downloaded; failure was a parsing bug (results record labels it fetch_failed).

Evidence quoted from the run

"atlases_stable_ge4_resolutions": 0

Adjudication note

Conditional on H1 holding, which it did not; Adams FDR fails at all resolutions (and resolution-scan log2FC ~-0.55 disagrees in sign with primary +0.98); AGTR1+ not found in Wang at any resolution.

Evidence quoted from the run

adams_2024,0.09759000729485331,0.7511151158040675,0.04879500364742666,0.8742219227185322,0.26911285945377744,0.373950482581387,0.34203517720054455,0.2526683107916884

Adjudication note

Descriptive; no layer significant in either atlas (all p>0.25); no dominance statement was written.

Evidence quoted from the run

The AGTR1+ cluster was not recoverable in either atlas (replication rate 0.0/2, 95% Wilson CI [0.0%, 65.8%], permutation p=1.0)

Adjudication note

Formal rule failed (report frames as 'pre-registered null') but attributes failure to DA yield: 52 and 27 DA cells, Adams 2 PD donors, Fullard SNpc absent. Strict reading would be NOT_SUPPORTED.

Evidence quoted from the run

H2 cannot be formally evaluated because no AGTR1+ subcluster was recovered in either atlas (Fig. 4).

Adjudication note

Only a TP53 proxy on all DA cells (FDR 0.190); NR2F2 gene set not cached. Report labels H2 'inconclusive'.

Evidence quoted from the run

The ROC AUC was 0.419 (95% bootstrap CI 0.217–0.644; Wilcoxon p=0.471; Fig. 5), falling below the pre-registered threshold of 0.75

Adjudication note

Panel derived from atlases with only 27-52 DA cells.

Evidence quoted from the run

(0 stable resolutions in both atlases; Fig. 2)

Adjudication note

Conditional on H1 passing; no atlas passed, so robustness of a non-existent effect is undefined.

Evidence quoted from the run

the dominant component of PD-vs-control DA-related change was component b — the DA-cluster fraction within total nuclei — with log2FC=−1.325 (95% CI −3.313 to −0.605)

Adjudication note

Descriptive; component b dominant in both atlases, component c CI spans zero; very low cell counts.

Evidence quoted from the run

No verbatim quote was recorded for this verdict.

Adjudication note

Run halted mid-execute (no active process; likely superseded by a relaunch) (Geneformer embedding still running); successive EXECUTE stages results/ directories empty; no results record. Hypotheses taken from latest plan the PLAN-stage hypothesis list.

Evidence quoted from the run

No verbatim quote was recorded for this verdict.

Adjudication note

Run halted mid-execute (no active process; likely superseded by a relaunch) (Geneformer embedding still running); successive EXECUTE stages results/ directories empty; no results record.

Evidence quoted from the run

No verbatim quote was recorded for this verdict.

Adjudication note

Run halted mid-execute (no active process; likely superseded by a relaunch) (Geneformer embedding still running); successive EXECUTE stages results/ directories empty; no results record.

Evidence quoted from the run

No verbatim quote was recorded for this verdict.

Adjudication note

Run halted mid-execute (no active process; likely superseded by a relaunch) (Geneformer embedding still running); successive EXECUTE stages results/ directories empty; no results record.

Evidence quoted from the run

No verbatim quote was recorded for this verdict.

Adjudication note

Run halted mid-execute (no active process; likely superseded by a relaunch) (Geneformer embedding still running); successive EXECUTE stages results/ directories empty; no results record.

Evidence quoted from the run

**H1 decision: NOT SUPPORTED.**

Adjudication note

GF AUC 0.775 vs baseline 0.749; dAUC +0.026 (CI -0.066 to +0.118, p=0.290). scGPT never run (package/weights missing, HF 401). Report itself notes ~19% power and 'more precisely described as inconclusive'.

Evidence quoted from the run

Inapplicable: no dopaminergic neurons in the Fullard atlas.

Adjudication note

Declared inapplicable already at plan stage.

Evidence quoted from the run

Deferred: GSE178265 (Kamath) count matrix absent. H3 inapplicable in this iteration.

Adjudication note

No adjudication note was recorded.

Evidence quoted from the run

Pre-registered subsampling levels (5K/25K/50K cells/donor) were not feasible on CPU/MPS hardware. Evaluated at 25/50/100 cells/donor

Adjudication note

Compute-limited proxy at 25/50/100 cells (50-500x below plan); run/report declare H4 'STABLE' (range 0.044) on the proxy levels, all dAUC CIs span 0. Could be read as SUPPORTED-on-proxy.

Evidence quoted from the run

| Geneformer (zero-shot) | 0.694 | [0.324, 1.000] | **MET** |

Adjudication note

DISCREPANCY: run/report say threshold MET, but pre-registered rule also requires CI lower bound >=0.50 (0.324 fails) on both atlases; Wang arm deferred (tar extraction pending); n=13 Adams. Baseline zero-shot AUC 0.944 exceeds GF. Strict reading would be NOT_SUPPORTED.

Evidence quoted from the run

H1: k=3, hypergeometric_p=0.3533, permutation_p=0.361, expected=2.11

Adjudication note

Module universe only 131, Fullard side mapped to 6 modules.

Evidence quoted from the run

"module_de_fdr": 0.995650997579099,

Adjudication note

Correlation arm implemented as module-vs-disease-label Spearman, not vs SCFA abundance.

Evidence quoted from the run

"note": "SNpc/V1 microglia not streamed in this run to conserve memory; DMNX only"

Adjudication note

Pipeline output 'SUPPORTED' (module FDR 0.011, z-corr perm p 0.012), but the 'cross-modality' correlation was computed between module score and disease label (code never uses LPS-producer signature) and spatial-specificity arm not run. Only DE component genuinely met. Re-run 093856 found FDR 0.514.

Evidence quoted from the run

H4: observed=0, expected=2.82, p=1

Adjudication note

Pre-registered Hallqvist 8-protein surrogate used.

Evidence quoted from the run

"h5_verdict": "all_below_threshold"

Adjudication note

Specificity control holds but is moot since H1 itself failed.

Evidence quoted from the run

Observed shared KEGG module count between K_w (33 Wallen-derived modules) and K_f (30 Fullard-derived modules) = 10, below the pre-registered threshold of 15.

Adjudication note

Hypergeometric p=0.175.

Evidence quoted from the run

**H2 NOT SUPPORTED.** BH-FDR = 0.463 > 0.05.

Adjudication note

No adjudication note was recorded.

Evidence quoted from the run

**H3 NOT SUPPORTED.** BH-FDR = 0.514 > 0.05.

Adjudication note

Cross-cohort correlation arm could not be formally tested (no shared donors).

Evidence quoted from the run

**H4 NOT SUPPORTED.** Permutation p = 0.373 > 0.05.

Adjudication note

No adjudication note was recorded.

Evidence quoted from the run

**H5 PASSED (all controls null).**

Adjudication note

All p=1.0; specificity control moot given H1 null.

Evidence quoted from the run

No verbatim quote was recorded for this verdict.

Adjudication note

Run halted mid-execute (no active process; likely superseded by a relaunch) while loading DMNX expression; EXECUTE-stage results empty, no results record.

Evidence quoted from the run

No verbatim quote was recorded for this verdict.

Adjudication note

Run halted mid-execute (no active process; likely superseded by a relaunch) while loading DMNX expression; EXECUTE-stage results empty, no results record.

Evidence quoted from the run

No verbatim quote was recorded for this verdict.

Adjudication note

Run halted mid-execute (no active process; likely superseded by a relaunch) while loading DMNX expression; EXECUTE-stage results empty, no results record.

Evidence quoted from the run

No verbatim quote was recorded for this verdict.

Adjudication note

Run halted mid-execute (no active process; likely superseded by a relaunch) while loading DMNX expression; EXECUTE-stage results empty, no results record.

Evidence quoted from the run

No verbatim quote was recorded for this verdict.

Adjudication note

Run halted mid-execute (no active process; likely superseded by a relaunch) while loading DMNX expression; EXECUTE-stage results empty, no results record.

Evidence quoted from the run

All effect sizes negative (PD < control). Primary count: 0/22 noncortical pairs (threshold >= 3).

Adjudication note

No adjudication note was recorded.

Evidence quoted from the run

Region-ordered Spearman rho = 0.500, one-sided permutation p = 0.825.

Adjudication note

No adjudication note was recorded.

Evidence quoted from the run

**Result: CORTEX SUPPORTED; SUBCORTEX NOT SUPPORTED**

Adjudication note

Cortical arm met (d=0.87-1.12); subcortical arm (DMNX FDR=0.216, GPi FDR=0.811) failed, so the AND rule is not fully met. Report notes EN>IN tested in PD donors only and may hold in controls too (not disease-specific).

Evidence quoted from the run

Max cortical vs subcortical Cohen's d difference: 0.171 (below 0.20 threshold).

Adjudication note

Executed test (correlation d-difference >0.20) deviates from the pre-registered LME 3-way interaction at FDR<0.05; results record lists sub-program genes with in_lamda_200=0.

Evidence quoted from the run

Orbitofrontal CosMx count matrix not present in Zenodo 10729767 deposit.

Adjudication note

No adjudication note was recorded.

Evidence quoted from the run

contract: missing: results record, analysis summary, figure legends, analysis code, summary, completion flag

Adjudication note

Run halted: successive EXECUTE stages produced no results (stage 4 log ends in a connection error); run later closed as halted and superseded by parkinson-lincs-reversal-r4. Hypotheses from PLAN stage 2 (latest plan).

Evidence quoted from the run

contract: missing: results record, analysis summary, figure legends, analysis code, summary, completion flag

Adjudication note

Run halted: successive EXECUTE stages produced no results (stage 4 log ends in a connection error); run later closed as halted and superseded by parkinson-lincs-reversal-r4. Hypotheses from PLAN stage 2 (latest plan).

Evidence quoted from the run

contract: missing: results record, analysis summary, figure legends, analysis code, summary, completion flag

Adjudication note

Run halted: successive EXECUTE stages produced no results (stage 4 log ends in a connection error); run later closed as halted and superseded by parkinson-lincs-reversal-r4. Hypotheses from PLAN stage 2 (latest plan).

Evidence quoted from the run

contract: missing: results record, analysis summary, figure legends, analysis code, summary, completion flag

Adjudication note

Run halted: successive EXECUTE stages produced no results (stage 4 log ends in a connection error); run later closed as halted and superseded by parkinson-lincs-reversal-r4. Hypotheses from PLAN stage 2 (latest plan).

Evidence quoted from the run

contract: missing: results record, analysis summary, figure legends, analysis code, summary, completion flag

Adjudication note

Run halted: successive EXECUTE stages produced no results (stage 4 log ends in a connection error); run later closed as halted and superseded by parkinson-lincs-reversal-r4. Hypotheses from PLAN stage 2 (latest plan).

Evidence quoted from the run

an uninterpretable null attributable to an underpowered evidence set (M = 33; minimum M ≈ 243 required for a testable H1)

Adjudication note

Formally fails the rule (k=0/50, q=1.0) and the run labels it 'NOT SUPPORTED'; but the evidence set was severely limited by CLUE API HTTP 401 (access) errors and H5 gate also failed (pre-reg: H1 not valid if H5 fails). Alternative reading: NOT_SUPPORTED.

Evidence quoted from the run

This result reflects a data-access limitation, not a biological finding, and H2 cannot be interpreted.

Adjudication note

Run labels 'NOT SUPPORTED', but MoA annotations were unavailable (CLUE API 401), so all classes had zero annotated members.

Evidence quoted from the run

The pre-registered threshold (r < 0.3 AND CI upper < 0.3) is satisfied: the 95% CI upper bound (0.172) is well below 0.3.

Adjudication note

r=0.133 [0.094,0.172]; computed on LINCS Level 2 data substituted for pre-registered Level 5 COMPZ; H5 sanity gate borderline-failed on same data.

Evidence quoted from the run

Neuronal-cell-line (NEU/NPC/FIBRNPC) top-50: k4 = 1; all-cell-lines top-50: k1 = 0. Fisher's exact p = 0.500

Adjudication note

Same M=33 evidence-set limitation as H1; could be read as INCONCLUSIVE.

Evidence quoted from the run

**H5 Decision: FAIL (borderline).**

Adjudication note

Ropinirole 47.8th percentile (<50th); pramipexole 53.6th; selegiline 87.2th.

Evidence quoted from the run

With n=2 PD, no meaningful bootstrap CI can be formed.

Adjudication note

Composite of H1a+H1b. Run labels H1 'NOT SUPPORTED' (ratio 1.000, CI [1.000,1.000]); Run executed with a 54-reaction minimal model, ATPM objective, and only 2 PD / 4 ctrl donors passing QC; review (stage 8) flagged ATPM objective as incapable of detecting glycolytic change and requested a re-execution, which (EXECUTE stage 9) did not produce the required outputs.

Evidence quoted from the run

the null primary result provides empirical evidence that minimal-model approaches with small sample sizes are insufficient for detecting cell-type-specific PD metabolic flux changes

Adjudication note

Observed ratio 1.000 [1.000,1.000] fails the rule literally (run: NOT SUPPORTED), but result is a model-degeneracy/n=2 artefact. Alternative reading: NOT_SUPPORTED.

Evidence quoted from the run

The directional labels tested by the binomial test are thus assigned to numerical noise rather than meaningful flux changes.

Adjudication note

Decision rule nominally met (15/18 = 83.3%, p=0.0038; run says CONFIRMED), but all glycolysis deviations are <=1e-11 (solver precision). Alternative reading: SUPPORTED (nominal).

Evidence quoted from the run

The p-value for 3/8 overlap against the 531-metabolite universe is 0.0083, but n_overlap = 3 < 4 required. H2 is not supported.

Adjudication note

TIMBR replaced by an iMAT exchange-flux surrogate (H2_SURROGATE).

Evidence quoted from the run

the inactivity is a model-architecture consequence, not a biological finding.

Adjudication note

All iron reactions ratio=0.000 (zero flux under ATPM objective); run labels NOT SUPPORTED. Alternative reading: NOT_SUPPORTED.

Evidence quoted from the run

The high ρ is consistent with both interpretations.

Adjudication note

Rule nominally met (iMAT vs E-Flux rho=0.944; run says CONFIRMED) but only 2 methods (no TIMBR) and report attributes high rho to bimodal model degeneracy. Alternative reading: SUPPORTED (nominal).

Evidence quoted from the run

"fullard_access": "not_attempted",

Adjudication note

Fullard atlas not attempted (no data-availability reason given); replaced by same-dataset PD vs PD+PDD sensitivity surrogate (H5_SURROGATE_da_only).

Evidence quoted from the run

**H1 is INCONCLUSIVE.** The n < 50 requirement is not met because the fetched supplements do not contain per-visit longitudinal group-mean time series

Adjudication note

rho=0.430 [-0.459,0.953], n=10 (FactorAnalysis fallback; GPy unavailable). Literal rule (CI lower <0.2 => not supported) would give NOT_SUPPORTED; data limitation could also justify NOT_TESTED_DATA.

Evidence quoted from the run

H2 is formally unevaluable under the pre-registration rules

Adjudication note

Leave-dMRI-out rho=0.349 at n=10, CIs not estimable. Could also be read as NOT_TESTED_DATA.

Evidence quoted from the run

**H3 verdict: NOT EVALUABLE.** Only one LRRK2 group-mean row was recovered.

Adjudication note

No adjudication note was recorded.

Evidence quoted from the run

**H4 verdict: VIOLATED — proteomics accounts for ~100% of the FA loading.**

Adjudication note

Report attributes dominance to feature-count imbalance (31 proteomics features) at n=10, not biology.

Evidence quoted from the run

Summary: 3 machine-readable (a), 2 PDF-table (b), 1 unreachable (c).

Adjudication note

Descriptive/classificatory hypothesis; 3/6 falls in pre-specified 'partial' band. Additionally 0/6 contained per-visit longitudinal group means.

Evidence quoted from the run

contract: missing: completion flag

Adjudication note

Run halted in PLAN (PLAN stage 3 did not complete; no completion flag); never reached EXECUTE. Hypotheses from the PLAN-stage hypothesis list (unfinalised plan).

Evidence quoted from the run

contract: missing: completion flag

Adjudication note

Run halted in PLAN (PLAN stage 3 did not complete; no completion flag); never reached EXECUTE. Hypotheses from the PLAN-stage hypothesis list (unfinalised plan).

Evidence quoted from the run

contract: missing: completion flag

Adjudication note

Run halted in PLAN (PLAN stage 3 did not complete; no completion flag); never reached EXECUTE. Hypotheses from the PLAN-stage hypothesis list (unfinalised plan).

Evidence quoted from the run

contract: missing: completion flag

Adjudication note

Run halted in PLAN (PLAN stage 3 did not complete; no completion flag); never reached EXECUTE. Hypotheses from the PLAN-stage hypothesis list (unfinalised plan).

Evidence quoted from the run

contract: missing: completion flag

Adjudication note

Run halted in PLAN (PLAN stage 3 did not complete; no completion flag); never reached EXECUTE. Hypotheses from the PLAN-stage hypothesis list (unfinalised plan).

Evidence quoted from the run

Strictly, H1 should be labelled NOT EVALUABLE rather than NOT SUPPORTED

Adjudication note

Run labels H1 'NOT SUPPORTED'; spatial (dbGaP-protected GeoMx) and proteomics (no processed table) unavailable; only a 1-modal snRNA surrogate was run (Factor 1 Wilcoxon p=0.948).

Evidence quoted from the run

Requires spatial modality per-donor count matrices (Δ R²_spatial between 3-modal and 2-modal MOFA+). No per-donor GeoMx spatial data could be obtained

Adjudication note

No adjudication note was recorded.

Evidence quoted from the run

If H1₀ is not rejected, H3 is "not evaluable."

Adjudication note

Discrepancy: run/report report H3 SUPPORTED (q_BH=5.07e-9) but on a 1-modal surrogate Factor 1 that captures DA cell-type composition, not the pre-registered H1 factor; per pre-reg H3 is not evaluable. Also k/OR differ between analysis summary (k=12, OR=10.6) and report (k=13, OR=11.7). Alternative reading: SUPPORTED (proxy).

Evidence quoted from the run

Requires all three modalities to compute per-modality variance fractions. With only snRNA-seq, the fraction is trivially 100% snRNA-seq.

Adjudication note

No adjudication note was recorded.

Evidence quoted from the run

Requires pairwise MOFA+ models with spatial and proteomics modalities.

Adjudication note

No adjudication note was recorded.

Evidence quoted from the run

**H1 (primary) BLOCKED**: DIA-NN 1.9.2 is Linux-only and unavailable on [the analysis platform]; PRIDE PXD046355 API returns empty file listing

Adjudication note

Blocked by raw-spectra inaccessibility and missing converter labels (also a tool/platform limitation).

Evidence quoted from the run

Hällqvist 2024 did not report Cox regression — the analysis used SVM and OPLS-DA, so there are no published Cox HRs to compare against.

Adjudication note

No adjudication note was recorded.

Evidence quoted from the run

Both H2 and H3 are blocked pending raw data access.

Adjudication note

No adjudication note was recorded.

Evidence quoted from the run

- **H4 is NOT SUPPORTED** (pre-registered threshold: ≥4/6 with p < 0.05)

Adjudication note

3/6 concordant, p=0.656. Hällqvist side used PD-vs-HC signed p-values (Figure_03) as a proxy for conversion HRs, which were not published; alternative reading NOT_TESTED_DATA.

Evidence quoted from the run

H5 requires per-sample protein intensities for the 72 iRBD samples to compute sub-panel AUCs, blocked by the same DIA-NN/PRIDE dependency as H1.

Adjudication note

No adjudication note was recorded.

Evidence quoted from the run

**Result: NOT SUPPORTED under the pre-registered label protocol. However, a post-hoc clean-negative sensitivity analysis recovers AUC = 0.925

Adjudication note

Primary AUC=0.554 [0.437,0.678] fails the rule (run: NOT SUPPORTED), but labels were inferred from MonthsSinceBaseline and contaminated (40 future converters in non-converter class); confirmed labels unavailable. Alternative reading: NOT_SUPPORTED.

Evidence quoted from the run

**Result: 6/8 concordant (75%) — SURROGATE ANALYSIS, not primary H2.**

Adjudication note

Reference Cox HRs not published; SVM coefficients used as surrogate (6/8, p=0.145, which would fail even loose acceptance). Pre-reg declared H2 'blocked' in this case. Alternative reading: NOT_SUPPORTED.

Evidence quoted from the run

The SDRF clinical covariates (age, sex, MoCA, UPSIT) are not available in the MOESM4 supplementary data.

Adjudication note

No adjudication note was recorded.

Evidence quoted from the run

**Result: NOT SUPPORTED (1/6 concordant, 17%).**

Adjudication note

Hällqvist-side HRs are the run's own Cox HRs from label-contaminated MOESM4 data (published conversion HRs unavailable).

Evidence quoted from the run

**Result: NOT SUPPORTED (no sub-panel exceeds AUC = 0.75).**

Adjudication note

Sub-panel AUCs 0.596/0.590/0.457; same label-contamination caveat as H1 applies.

Evidence quoted from the run

"reasons": "contract: missing: results record, analysis summary, figure legends, analysis code, summary, completion flag"

Adjudication note

Run halted at EXECUTE stage 3 (required stage outputs missing; no results record in successive EXECUTE stages); hypothesis never tested.

Evidence quoted from the run

"reasons": "contract: missing: results record, analysis summary, figure legends, analysis code, summary, completion flag"

Adjudication note

Run halted at EXECUTE stage 3 (required stage outputs missing; no results record in successive EXECUTE stages); hypothesis never tested.

Evidence quoted from the run

"reasons": "contract: missing: results record, analysis summary, figure legends, analysis code, summary, completion flag"

Adjudication note

Run halted at EXECUTE stage 3 (required stage outputs missing; no results record in successive EXECUTE stages); hypothesis never tested.

Evidence quoted from the run

"reasons": "contract: missing: results record, analysis summary, figure legends, analysis code, summary, completion flag"

Adjudication note

Run halted at EXECUTE stage 3 (required stage outputs missing; no results record in successive EXECUTE stages); hypothesis never tested.

Evidence quoted from the run

"reasons": "contract: missing: results record, analysis summary, figure legends, analysis code, summary, completion flag"

Adjudication note

Run halted at EXECUTE stage 3 (required stage outputs missing; no results record in successive EXECUTE stages); hypothesis never tested.

Evidence quoted from the run

[EUR] R²_obs=0.062550 | R²_L=0.026740 | AUC=0.5467 | 95% CI=[0.5440, 0.5501]

Adjudication note

Latest stage EXECUTE stage 9 (liftover + all-chrom AFs fixed) gives AUC 0.547 vs benchmark 0.621 (deviation_auc -0.074349); EXECUTE stage 4 and reviewed report also REJECTED (AUC 0.548, -0.073). Stage 9 lacks analysis summary/completion flag (stage incomplete) but results record is complete.

Evidence quoted from the run

"ratio_eas_over_eur": 0.992253,

Adjudication note

Rule met formally (deviation +0.061 within ±0.10), but ratio CI [0.984, 1.000] excludes 0.931 and EAS<EUR ordering is not resolved; in stage 4 (basis of the reviewed report) support was explicitly 'artefactual' (build mismatch). stage 9 fixed liftover but was never analysed/reviewed. Low-confidence support of a weak tolerance test.

Evidence quoted from the run

The Storm 2025 (Brain 148:2412) supplementary pQTL-instrumented gene list is not available in `storm_2025_pd_targets.json` (empty array). H3 cannot be evaluated.

Adjudication note

DEFERRED in stages 4 and 9 (recorded H3 verdict: DEFERRED).

Evidence quoted from the run

[EUR] R²_obs=0.092491 | R²_L=0.039539 | AUC=0.5570 | 95% CI=[0.5539, 0.5608]

Adjudication note

DISCREPANCY: stage 9 code labels H4 SUPPORTED, but with corrected Nalls N_eff (145177) Nalls AUC 0.557 > Kim 0.547; results record delta_kim2024_vs_nalls2019 = -0.01032, CI [-0.014894, -0.005571] (Nalls significantly better) -> pre-registered REJECTED condition; reported delong_p=1.0 is the Kim>Nalls direction. Reviewed report (stage 4 data) said REVERSED, driven by an N_eff artefact. Analytical (label-free) AUCs; interpret cautiously.

Evidence quoted from the run

H5 is rejected — not because uncertainty is genuinely small, but because the chi-squared parametric bootstrap models uncertainty in a chi-squared sampling model

Adjudication note

Formal rule outcome REJECTED in stages 4 and 9 (AAC CI width ~0.011-0.014), but the run itself states the bootstrap measures analytical sensitivity, not sampling uncertainty (no individual-level labelled genotypes), so the test cannot answer the question. Could also be classed NOT_SUPPORTED by the letter of the rule.

Evidence quoted from the run

3/20 TFs (15%) show sign agreement in both — no better than a null of 50% sign-agreement per pair (binomial p = 0.785)

Adjudication note

Pre-registered design not executable: Fullard SNpc discovery atlas unavailable (Adams n=4 PD used), CollecTRI/ULM instead of pySCENIC/AUCell, 2 of 3 replication atlases. Chance-level result (p vs shuffled 0.919); run calls it 'inconclusive rather than falsified' / underpowered. stage 4 (basis of the report) gave 9/20 in 1 atlas, p=0.748. Arguably NOT_TESTED_DATA (proxy test).

Evidence quoted from the run

This is a proxy test, not the pre-registered test. Conclusion about H2 is inconclusive pending the actual Kim 2024 supplementary table.

Adjudication note

Kim 2024 fine-mapping table not retrieved; OpenTargets proxy gave 0 overlaps, q=1.0 (run labelled NOT_SUPPORTED).

Evidence quoted from the run

INAPPLICABLE. Prior workspace `parkinson-da-vulnerability` AUCell scores not available in this execution environment.

Adjudication note

No adjudication note was recorded.

Evidence quoted from the run

Mean AUC = 0.808 meets the ≥ 0.70 threshold, but the pre-registration criterion required CI lower bound > 0.5. With only 2 LOO folds, bootstrap CI lower bound = 0.25 (< 0.5).

Adjudication note

Run labels NOT_SUPPORTED; 2 folds/16 donors instead of 4-fold LOO, CI [0.25,1.0] uninformative (run attributes to data limitation). Could be NOT_SUPPORTED by the letter of the rule.

Evidence quoted from the run

SUPPORTED. Swing = 1 ≤ pre-registered threshold of 1.

Adjudication note

Robustness of a chance-level count (2-3/20); report (stage 4 data) reported swing=0.

Evidence quoted from the run

The primary hypothesis is not evaluable with the data currently in the shared cache.

Adjudication note

Run labelled NOT SUPPORTED (0/1 atlases), but only Fullard tested and its leukocytes were DMNX-dominated with zero SNpc cells; Kamath/Wang/Adams not downloaded (file size/run budget). Arguably NOT_TESTED_DATA.

Evidence quoted from the run

ρ = 0.500, permutation p = 0.505.

Adjudication note

Run labelled NOT SUPPORTED; SNpc absent (0 leukocytes), only 3 regions so p<0.05 is unreachable.

Evidence quoted from the run

The Census query fetched only leukocyte-annotated cells; dopaminergic neurons (AGTR1+ DA cluster) were not downloaded.

Adjudication note

DEFERRED; data not fetched (could be classed NOT_TESTED_OTHER).

Evidence quoted from the run

GSE243639 (Martirosyan 2024 Visium data) RAW.tar (1.2 GB) was not downloaded.

Adjudication note

DEFERRED; the sibling run later found GSE243639 is snRNA-seq, not Visium.

Evidence quoted from the run

the exhaustion module (PDCD1, HAVCR2, LAG3, TOX, TIGIT) shows mean PD − ctrl difference of +0.012, Wilcoxon padj = 1.000.

Adjudication note

Run claims SUPPORTED, but the rule needs >=3/4 atlases and only 1 (wrong-tissue) atlas was tested.

Evidence quoted from the run

Zero of two testable atlases (Wang 2024 and Fullard 2024) show both log2FC > 0 and FDR < 0.1

Adjudication note

Also 0/3 with Martirosyan extension; Fullard (62 PD/17 ctrl) null, Cohen d -0.30. Kamath not tested (matrix not in cache), Adams excluded (1 PD donor).

Evidence quoted from the run

The pre-registered rank order (GPi > M1 > PFC > DMNX > V1) does not match the observed median scores (ρ = −0.10, p = 0.60).

Adjudication note

No adjudication note was recorded.

Evidence quoted from the run

No atlas meets the pre-registered criterion of ρ > 0.3 and permutation p < 0.05.

Adjudication note

Very small donor n (5, 5, 15); Fullard rho -0.40 (opposite direction). Somewhat underpowered.

Evidence quoted from the run

The Martirosyan 2024 dataset (GSE243639) is snRNA-seq, not Visium spatial transcriptomics.

Adjudication note

Ma 2025 Visium likely dbGaP-restricted; deferred at PLAN.

Evidence quoted from the run

The failure of H5 to show a false-positive enrichment is expected (and reassuring), but the failure of H1 means the negative control comparison is moot — both modules are null.

Adjudication note

Rule met on 3 atlases (Wang, Fullard, Martirosyan; results record H5 exhaustion replication rate 0.0) but uninformative because H1 is null.

Evidence quoted from the run

(direction): Mean log2FC = +0.4846 (95% CI +0.182 to +0.787), 8/9 subunits individually up, permutation p = 0.167 at its achievable floor.

Adjudication note

Pre-registered direction rule met at n=2+2 (no formal significance possible). Knowledge-graph nodes = supported. CAVEAT: Δ vs prior = -0.0004; report itself questions whether GSE167335 is independent of the original cohort (possible re-analysis of the same data).

Evidence quoted from the run

E-MTAB-7303 (pre-registered H2 cohort) remains BLOCKED — raw FASTQ only, no alignment tools installed.

Adjudication note

Only proxy-tested post hoc in GSE62642 (n=4 vs 2): module d = -0.042, 95% CI [-2.531, 1.563], NULL/underpowered. Knowledge-graph nodes = inconclusive. If the proxy is accepted, verdict would be INCONCLUSIVE (underpowered null).

Evidence quoted from the run

AUC = 1.000 with CI = [1.0, 1.0] is uninformative due to complete separation at this sample size.

Adjudication note

Letter of the rule met (d=+5.14, CI [5.142, 15.558]; AUC CI [1,1]) and graph status 'supported', but CIs are degenerate at n=2+2, perm p=0.167; run states formal inference is impossible.

Evidence quoted from the run

DA vs non-DA contrast: Δd = +4.065, 95% CI [−174.663, +15.394], p = 0.741

Adjudication note

Run verdict GLOBAL/inconclusive; pre-registered rule would call this NULL; n=2+2 per cell type.

Evidence quoted from the run

No cohort in this study can achieve formal statistical significance at α = 0.05 in the available data.

Adjudication note

Descriptive sub-analyses with no formal decision rule. H1 direction robust to normalisation (+0.485 vs +0.490); gene-set split not robust (V0-only down -1.35 in GSE157783 while 9-gene up); no cohort adequately powered.