Campaign run · Parkinson's disease

parkinson-multimodal-spatial-proteomics-r3

PDCompleted the pipelineAttempt 3 of 3 at this topic

Does a MOFA+ (Multi-Omics-Factor-Analysis) joint embedding of three modalities of PD substantia-nigra data — Martirosyan 2024 snRNA-seq + Visium (GSE243639), Ma 2025 GeoMx spatial (PMC12322057), and Zhu 2024 LC-MS/MS DLPFC proteomics (PRIDE PXD030142) — reveal factor(s) that ONLY appear when all three modalities are jointly modelled, and that are unrecoverable from any pairwise integration?

Not testable (data) 5

Open the final report (PDF)This run in the hypothesis explorer

Status

How far it got: Completed the pipeline
Tested at least one hypothesis and reached the multi-agent review stage. This shows the run went through the whole pipeline, not that its work passed a quality bar.
Hypotheses: 5 nominated; 5 evaluated.
Recorded status: Halted. The run recorded a final status of halted: it was closed without being marked done. A halted run can still have tested hypotheses and written a final report.
Census class: Success. Census class is a label from an earlier, process-level census of the campaign. It does not map one-to-one onto the stages above. How the classes compare.

Research question

The sentence under the run id is quoted from this brief.

Full research brief given to the run

Multi-modal integration has become a default expectation in PD research, but the value-add of going from 2-modal to 3-modal is rarely tested with a clean ablation. Pairwise integrations in PD exist: snRNA × Visium (Martirosyan 2024); snRNA × proteomics (Zhu 2024); spatial × CosMx (Goralski 2024). Joint 3-modal MOFA+ on PD-SN has not been published. This OmiCairn run pre-registers the ablation that asks whether the third modality is necessary or decorative.

Hypotheses (pre-registered):

H1 (primary — all-modalities factor exists): ≥ 1 MOFA+ factor explains > 5% variance in EACH of the three modalities simultaneously, AND is associated with PD diagnosis (per-donor-group Wilcoxon p < 0.05). Pre-registered name: "all-modalities factor".

H2 (secondary — proteomics necessity ablation): Removing the proteomics modality and re-fitting MOFA+ produces a closest-matching factor (max absolute loading correlation with the H1 factor) whose explained variance in the spatial modality drops by ≥ 30 percentage points. Proteomics is NECESSARY for the joint factor, not nice-to-have.

H3 (secondary — top-loaded gene enrichment): Top-100 loaded genes of the all-modalities factor are enriched (q < 0.05 hypergeometric) for the union of: Goralski 2024 LAMDA cortical signature genes, AND Hemedan 2024 Boolean-PD "key-regulator" gene list.

H4 (secondary — modality-balance check): No single modality contributes > 60% of the explained variance summed across the H1 factor (otherwise the "all-modalities" label is misleading).

H5 (secondary — pairwise comparison): Repeat the analysis with three pairwise MOFA+ models — (snRNA + Visium), (snRNA + proteomics), (Visium + proteomics). Test whether any pairwise model recovers a factor with ≥ 0.7 absolute loading correlation to the H1 3-modal factor. If yes, the 3-modal claim is weaker than headline. If no, 3-modal is necessary.

Null payoff. A null all-modalities factor (H1: no factor >5% variance in all three modalities, or no PD association) means PD SN biology is dominated by the strongest single modality (probably snRNA-seq) and multi-omics doesn't add. Canonical ablation answer the field has been asking.

The research question comes from the topic's current research brief. All attempts at a topic share it; earlier attempts may have started from an earlier wording.

Hypotheses and verdicts

Each hypothesis was written with its decision rule before any analysis (or, for on-the-fly hypotheses, recorded during execution). AI adjudication agents applied that rule to the run's latest outputs and quoted the evidence verbatim. Where the data did not meet the rule, the verdict follows the data, even if the run's own label or report claimed support.

Each hypothesis as pre-registered (a condensed statement with its decision rule), the adjudicated verdict, the verbatim evidence and the adjudicator's note.
ID	Hypothesis and decision rule	Verdict
H1Pre-registered	3-modal MOFA+ (snRNA+spatial+proteomics) yields >=1 factor with R2>5% in all three modalities AND PD-associated (Wilcoxon p<0.05).	Not testableIn the explorer
Evidence (verbatim from the run outputs) 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).
H2Pre-registered	Proteomics ablation drops spatial R2 of matched factor by >=30 pp.	Not testableIn the explorer
Evidence (verbatim from the run outputs) 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 None recorded.
H3Pre-registered	Top-100 H1-factor genes enriched for LAMDA ∪ Hemedan (hypergeometric q<0.05); not evaluable if H1 null not rejected.	Not testableIn the explorer
Evidence (verbatim from the run outputs) 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).
H4Pre-registered	No single modality contributes >60% of H1-factor explained variance.	Not testableIn the explorer
Evidence (verbatim from the run outputs) Requires all three modalities to compute per-modality variance fractions. With only snRNA-seq, the fraction is trivially 100% snRNA-seq. Adjudication note None recorded.
H5Pre-registered	No pairwise MOFA+ model recovers factor with \|r\|>=0.70 to the H1 factor.	Not testableIn the explorer
Evidence (verbatim from the run outputs) Requires pairwise MOFA+ models with spatial and proteomics modalities. Adjudication note None recorded.

Final report

Title (as written by the system)

3MAP: a pre-registered three-modal MOFA+ audit protocol for Parkinson's disease substantia nigra reveals data-access barriers and an 11.7-fold LAMDA-enriched dopaminergic factor

Main conclusion

Spatial and proteomics per-donor matrices could not be accessed, so only an snRNA-seq surrogate was run. Factor 1 explains 26.5% of variance but has no PD association (Wilcoxon p=0.948; H1 not supported). It is 11.7-fold enriched for LAMDA∪Hemedan PD genes (q_BH=5.07e-9; H3 supported). H2, H4 and H5 could not be evaluated.

Report titles are the system's own and can state a result more strongly than the report's verdicts support. Read the main conclusion and the verdicts with it. The main conclusion was written during release review and keeps the report's own numbers and verdicts.

The PDF is published exactly as released. It may call itself a “paper”, call the system's internal multi-agent review “peer review”, and cite internal stage folders by number.

Open the final report (PDF, 492 KB)

All 3 attempts at this topic

parkinson-multimodal-spatial-proteomics-r1Planned, no test completed
parkinson-multimodal-spatial-proteomics-r2Stopped before planning
parkinson-multimodal-spatial-proteomics-r3Completed the pipeline Report

Attempts are numbered in launch order. A topic was relaunched when an earlier attempt stopped or was superseded.