Perturbation Cascade Network · Stage C1
Real CRISPRi responses for 21 candidates seed a 1,360-node Reactome network. Topology propagation, competitive pathway support, heterogeneity and DMD projection are exposed as separate evidence states, so a coupled pathway hypothesis can be explored without being mislabeled as a measured DMD candidate response.
Candidate pathway cascades · PCN-v0.1
| Candidate | Source cells | Pathway split-half ρ | Graph CV ρ | Permuted CV ρ | DMD rescue projection ρ | Correction alignment ρ | Decision |
|---|---|---|---|---|---|---|---|
| ADAM10 | 161 | 0.818 | 0.860 | 0.113 | 0.331 | 0.082 | Prospective DMD Perturb-seq |
| CALR | 86 | 0.414 | 0.766 | 0.059 | 0.193 | 0.123 | Prospective DMD Perturb-seq |
| MPHOSPH6 | 83 | 0.752 | 0.857 | 0.004 | 0.151 | 0.189 | Prospective DMD Perturb-seq |
Candidate-level pathway changes are now measured in the source perturbation assay and propagated through an explicit network. The cascade can reveal coordinated downstream modules instead of a single-gene score.
The three priorities are cross-context hypotheses. The other 18 candidates retain explicit abstention reasons, and none of the 21 has yet returned an independent DMD-muscle perturbation outcome.
Candidate statistical bridge · PCB-v0.2
The release now separates raw differential-response burden, competitive Reactome support, deterministic direction stability and source-to-source heterogeneity. The network is therefore auditable beyond a visual pathway graph.
Cells are not biological replicates, pathways overlap and the DMD source axes are not independent trials. These calculations prioritize falsifiable experiments; they do not convert a cross-context projection into a virtual-cell outcome.
Observed subnetwork · GSE293514
Grouped CROP-seq: lowest myogenesis differentiation score group
Median member LFC 1.52. This is a descriptive convergence summary, not a new module-level test.
Grouped CROP-seq: no exact direction encoded
Median member LFC 0.704. This is a descriptive convergence summary, not a new module-level test.
Grouped CROP-seq: no exact direction encoded
Median member LFC 1.179. This is a descriptive convergence summary, not a new module-level test.
Grouped CROP-seq: no exact direction encoded
Median member LFC 2.042. This is a descriptive convergence summary, not a new module-level test.
Grouped CROP-seq: no exact direction encoded
Median member LFC 2.003. This is a descriptive convergence summary, not a new module-level test.
Grouped CROP-seq: no exact direction encoded
Median member LFC 0.921. This is a descriptive convergence summary, not a new module-level test.
Grouped CROP-seq: no exact direction encoded
Median member LFC 0.688. This is a descriptive convergence summary, not a new module-level test.
Grouped CROP-seq: no exact direction encoded
Median member LFC 0.697. This is a descriptive convergence summary, not a new module-level test.
Grouped CROP-seq: no exact direction encoded
Median member LFC 1.627. This is a descriptive convergence summary, not a new module-level test.
Grouped CROP-seq: no exact direction encoded
Median member LFC 0.626. This is a descriptive convergence summary, not a new module-level test.
Grouped CROP-seq: highest myogenesis differentiation score group
Median member LFC 1.076. This is a descriptive convergence summary, not a new module-level test.
Subset boundary: the publication reports 160 hits in 23 complexes and 1,286 PPIs. The article text names only a subset in machine-resolvable form, so this release encodes ten named complexes plus MYMK and does not invent the missing memberships.
State signatures
MYOG · ACTA1 · ERBB3 · ACTN2 · SRL · UNC45B · APOBEC2 · CDKN1C
observed unperturbed reference not perturbation deltaRRM2 · BIRC5 · CENPF · UBE2C · LMNB1 · ZWINT · HMGA1 · CKS2
observed unperturbed reference not perturbation deltaCHRNG · ARPP21 · ACTA1 · TNNI1 · CACNA1S · MYBPH · CASQ2 · LDB3
observed signature definition not candidate deltaMT2A · SERPINE1 · NPTX2 · MKI67 · TOP2A · HMGA2 · CENPF · NT5E
observed signature definition not candidate deltaThe matrix has 62,711 expression columns, while the recoverable public gene order has 62,710 rows. Do not attach gene symbols to the public CROP-seq matrix and do not emit candidate-level pathway deltas from it.
Current candidate audit
Twelve candidates were not in the MyoCRISPR library. A non-hit is specific to this healthy-myoblast fusion endpoint and must not be interpreted as “no DMD effect.”
| Gene | In library | Screen state | Fusion LFC | FDR | Interpretation |
|---|---|---|---|---|---|
| ADAM10 | Yes | Assessed | -0.0752 | 1 | ASSESSED NO FUSION HIT AT FDR 0 1 |
| CALR | Yes | Assessed | -0.0659 | 1 | ASSESSED NO FUSION HIT AT FDR 0 1 |
| CPEB1 | No | Not assessed | — | — | NOT INCLUDED IN MYOCRISPR LIBRARY |
| DDX19B | Yes | Assessed | 0.0691 | 1 | ASSESSED NO FUSION HIT AT FDR 0 1 |
| DNAAF3 | No | Not assessed | — | — | NOT INCLUDED IN MYOCRISPR LIBRARY |
| DNM1 | No | Not assessed | — | — | NOT INCLUDED IN MYOCRISPR LIBRARY |
| EHMT2 | No | Not assessed | — | — | NOT INCLUDED IN MYOCRISPR LIBRARY |
| EPS8L1 | No | Not assessed | — | — | NOT INCLUDED IN MYOCRISPR LIBRARY |
| GFOD2 | No | Not assessed | — | — | NOT INCLUDED IN MYOCRISPR LIBRARY |
| INTS13 | No | Not assessed | — | — | NOT INCLUDED IN MYOCRISPR LIBRARY |
| LMO2 | No | Not assessed | — | — | NOT INCLUDED IN MYOCRISPR LIBRARY |
| MON1A | Yes | Assessed | -0.107 | 1 | ASSESSED NO FUSION HIT AT FDR 0 1 |
| MPHOSPH6 | Yes | Assessed | 0.372 | 0.779 | ASSESSED NO FUSION HIT AT FDR 0 1 |
| NAGLU | Yes | Assessed | 0.0611 | 0.685 | ASSESSED NO FUSION HIT AT FDR 0 1 |
| RAC3 | No | Not assessed | — | — | NOT INCLUDED IN MYOCRISPR LIBRARY |
| RNASEH2C | Yes | Assessed | 0.18 | 0.685 | ASSESSED NO FUSION HIT AT FDR 0 1 |
| RNF8 | No | Not assessed | — | — | NOT INCLUDED IN MYOCRISPR LIBRARY |
| WDR4 | Yes | Assessed | -0.0103 | 1 | ASSESSED NO FUSION HIT AT FDR 0 1 |
| ZFP69B | No | Not assessed | — | — | NOT INCLUDED IN MYOCRISPR LIBRARY |
| ZNF133 | Yes | Assessed | -0.0035 | 1 | ASSESSED NO FUSION HIT AT FDR 0 1 |
| ZNF236 | No | Not assessed | — | — | NOT INCLUDED IN MYOCRISPR LIBRARY |
DMD hypothesis layer
Candidate-level intervention identity; no DMD response is inferred from identity alone.
Observed same-assay or source-linked molecular response substrate.
Mechanism modules used to organize hypotheses across DMD source-state and target-state evidence.
Myogenic and disease-state compartments where a future perturbation-conditioned model would need truth.
Assay readouts required before a disease-relevant effect can be claimed.
Decision-blocking evidence that is still absent in the frozen release.
Layout boundary: Node position is an explanatory layout artifact; only node identity, edge direction and evidence_state carry scientific meaning.
Functional validation scenarios
Evidence boundary: Stage C1 emits measured HepG2 pathway deltas and bounded topology predictions. These historical design scenarios identify the DMD molecular and functional outcomes still required; measured_dmd_pathway_delta, cell_state_delta and functional_effect remain absent.
Direction: direction not predicted current release
Measure together: fusion index · myotube marker panel · viability
Matched DMD/control myogenic perturbation with prespecified fusion and toxicity endpoints.Direction: direction not predicted current release
Measure together: myotube area · nuclei per myotube · cell survival
Prospective effect margin and null/toxic interpretation frozen before outcomes are visible.Current interpretation: ZNF133 has a real healthy-myoblast screen result, but it is an endpoint-specific non-hit. The result weakens a generic fusion-defect claim without establishing absence of a DMD-context effect.
Next experiment: Only retain ZNF133 in a matched DMD/control pilot if its DMD-specific rationale is prespecified; measure target engagement, early transcriptome, fusion and viability together.
Direction: direction not predicted current release
Measure together: oxygen consumption · atp proxy · calcium handling
Audit external evidence, then measure energy and calcium endpoints in the same DMD/control perturbation study.Direction: direction not predicted current release
Measure together: calcium transient · contractile force · viability
Late myotube function assay with toxicity gate and independent replication.Current interpretation: GFOD2 is represented as an energy-function propagation hypothesis, but external evidence is still audit-pending and no functional outcome exists.
Next experiment: Resolve the evidence audit first, then pair energy-state and calcium/contractile readouts in a registered DMD myotube study.
Direction: direction not predicted current release
Measure together: early transcriptome · target engagement · state marker panel
Validate regulator-target bridge in the declared myogenic cell state.Direction: direction not predicted current release
Measure together: cell survival · differentiation capacity · reagent concordance
Safety-first perturbation study with two reagents before interpreting downstream state change.Current interpretation: CPEB1 is a safety-first propagation scenario: early molecular and regulatory context must be separated from toxicity or differentiation blockade.
Next experiment: Measure target engagement, early transcriptome, viability and differentiation capacity before assigning pathway meaning.
System coupling
Exact healthy-human-myoblast LFC and FDR values anchor the functional network.
Orthogonal gene-level validation is retained as a qualitative direction when no per-gene effect is published.
DMD source-state, myogenic reference and regulatory priors organize possible propagation routes.
These edges point to the exact assays needed before any disease-response claim.
Focal cascades
The healthy-myoblast screen is an observed endpoint-specific non-hit; the DMD response remains unmeasured.
The non-hit constrains one healthy-myoblast fusion hypothesis but neither predicts nor excludes a DMD effect.Mechanism module remains audit pending and requires late functional truth.
External or contextual evidence cannot create a DMD function prediction.Safety and state-transition endpoints must be measured before interpreting mechanism.
The cascade records what would need to move together; it does not say that it will.DMD hypothesis edge audit
| Arrow | Meaning | Evidence state | Basis | Unlock condition |
|---|---|---|---|---|
| PERT:ZNF133 → ASSAY:HEPG2_CRISPRI | observed response substrate | observed same assay | Processed HepG2 CRISPRi response vector exists for ZNF133. | None for same-assay evidence; DMD transfer still locked. |
| PERT:GFOD2 → ASSAY:HEPG2_CRISPRI | observed response substrate | observed same assay | Processed HepG2 CRISPRi response vector exists for GFOD2. | Audit pending external evidence before changing candidate status. |
| PERT:CPEB1 → ASSAY:HEPG2_CRISPRI | observed response substrate | observed same assay | Processed same-assay response can seed a candidate hypothesis. | Run matched DMD/control myogenic perturbation before disease interpretation. |
| PERT:ZNF133 → ASSAY:MYOBLAST_CONTEXT_SCREEN | screened; no fusion hit | assessed no hit | ZNF133 was assessed in GSE293514 (positive-selection LFC -0.003533; FDR 0.999999) and did not meet the fusion-hit threshold. | Do not infer no DMD effect; test a prespecified endpoint in matched DMD/control myogenic cells. |
| ASSAY:HEPG2_CRISPRI → PROGRAM:DMD_SOURCE_STATE | cross-context comparison | contextual bridge | Same-assay response is compared against DMD source-linked programs only as context. | Establish transfer on a disease-relevant perturbation holdout. |
| PROGRAM:DMD_SOURCE_STATE → PROGRAM:NICHENET_TARGET_STATE | regulatory target-state bridge | inferred bridge | NicheNet target-state evidence supplies regulatory context. | Validate ligand-target or regulator-target mechanism in the declared cell context. |
| PROGRAM:DMD_SOURCE_STATE → PATHWAY:ECM_FIBROSIS | disease module placement | hypothesis | DMD source-state evidence motivates an ECM/fibrosis module. | Measure ECM or fibrosis readouts after candidate perturbation in DMD-relevant cells. |
| PROGRAM:NICHENET_TARGET_STATE → PATHWAY:IMMUNE_CROSSTALK | cross-cell regulatory hypothesis | hypothesis | Regulatory context motivates a cell-cell consequence module. | Test causal cross-cell outcome or coculture readout. |
| ASSAY:MYOBLAST_CONTEXT_SCREEN → PATHWAY:MYOGENIC_DIFFERENTIATION | myogenic bridge | contextual bridge | Fusion-screen context links candidate selection to myogenic biology. | Repeat with disease and matched-control perturbation outcome. |
| PATHWAY:MYOGENIC_DIFFERENTIATION → STATE:PROLIFERATING_MYOBLAST | reference start state | contextual bridge | Unperturbed human-myogenic trajectory anchors the starting compartment. | Add perturbation-conditioned cells at 0 h and early response times. |
| PATHWAY:MYOGENIC_DIFFERENTIATION → STATE:FUSING_MYOCYTE | transition context | contextual bridge | The platform can place a question on the myoblast-to-fusion axis. | Measure candidate-specific state transition in DMD and control cells. |
| PATHWAY:MYOGENIC_DIFFERENTIATION → STATE:MATURING_MYOTUBE | late state context | contextual bridge | The unperturbed trajectory includes a late reference state. | Add perturbation-conditioned myotube maturation outcomes. |
| PATHWAY:MITO_ENERGY → FUNCTION:CALCIUM_CONTRACTION | function-coupled hypothesis | hypothesis | Energy-state changes would need to be tied to contractile readouts. | Measure calcium handling and contraction under candidate perturbation. |
| PATHWAY:ECM_FIBROSIS → FUNCTION:MEMBRANE_INTEGRITY | matrix-function hypothesis | hypothesis | ECM remodeling could matter only if functional readouts change. | Measure membrane injury or repair endpoint in the same experiment. |
| STATE:FUSING_MYOCYTE → FUNCTION:FUSION_INDEX | state-to-endpoint mapping | contextual bridge | Fusion index is an interpretable endpoint for a myogenic transition. | Freeze endpoint, effect margin and analysis before outcomes are visible. |
| STATE:MATURING_MYOTUBE → FUNCTION:CALCIUM_CONTRACTION | late functional endpoint | hypothesis | Mature myotube state needs functional measurement before disease interpretation. | Run late myotube function assay with viability co-readout. |
| STATE:PROLIFERATING_MYOBLAST → FUNCTION:VIABILITY_TOXICITY | early safety gate | hypothesis | Any apparent mechanism is uninterpretable without viability/toxicity checks. | Include toxicity controls and reagent concordance in the pilot. |
| FUNCTION:FUSION_INDEX → GAP:DMD_PERTURBATION_TRUTH | requires DMD truth | missing validation | No candidate perturbation fusion outcome exists in matched DMD/control cells. | Generate a registered DMD/control myogenic perturbation outcome. |
| FUNCTION:MEMBRANE_INTEGRITY → GAP:DMD_PERTURBATION_TRUTH | requires disease functional truth | missing validation | No membrane integrity endpoint is measured for current candidate perturbations. | Add disease-relevant functional endpoint and analysis threshold. |
| FUNCTION:CALCIUM_CONTRACTION → GAP:DMD_PERTURBATION_TRUTH | requires muscle function truth | missing validation | No calcium or contraction outcome exists for the candidate perturbations. | Measure late function in a matched, preregistered assay. |
| GAP:DMD_PERTURBATION_TRUTH → GAP:PROSPECTIVE_REPLICATION | must replicate before claim | missing validation | The release has zero prospective predictions and zero returned outcomes. | Register prediction, run blinded outcome, then replicate independently. |
| ASSAY:MYOBLAST_CONTEXT_SCREEN → GAP:DMD_PERTURBATION_TRUTH | healthy context cannot substitute DMD truth | missing validation | The observed GSE293514 fusion endpoint is from healthy human myoblasts and is not a matched DMD perturbation outcome. | Generate a registered, donor-aware DMD/control perturbation study with molecular and functional readouts. |
Claim boundary
The release contains measured HepG2 candidate pathway deltas, Reactome topology cascade predictions, measured DMD induction/correction reference axes and cross-context DMD projections. It does not contain a measured candidate perturbation in DMD muscle, a calibrated treatment probability, functional rescue or a therapeutic recommendation.
{
"network_schema": "nmd-vcell-perturbation-cascade-network/1.5",
"analysis_release": "PCN-v0.3-20260804",
"lifecycle_state": "MEASURED_SOURCE_RESPONSE_PLUS_BOUNDED_DMD_PROJECTION",
"observed_human_myoblast_edges": 92,
"measured_hepg2_candidate_responses": 21,
"reactome_pathway_nodes": 1360,
"reactome_topology_edges": 11336,
"candidate_level_hepg2_pathway_delta_emitted": true,
"topology_cascade_prediction_emitted": true,
"cross_context_dmd_projection_emitted": true,
"reactome_competitive_tests": 28938,
"split_half_direction_robust_candidate_pathway_pairs": 1782,
"projected_dmd_reversal_rows": 3210,
"provisional_dmd_perturb_seq_priorities": 3,
"measured_dmd_candidate_replications": 0,
"validated_dmd_prediction_edges": 0
}