PIGA · A7SEM · A7SEM Reverse | Empirical research pathway
Evidence before inference. Verification before action.
A transparent, local-first workspace for exploring public cancer research catalogues, combining aggregate multi-omics summaries and designing falsifiable research validations. Not a clinical decision tool, not a diagnostic algorithm.
Start here · Upload or enter evidence ↗5 omics modalitiesGDC + cBioPortal metadataBaseline comparisonWet-lab / prospective protocolsClinical Gate 2: BLOCKFive clinical evidence workstreams · not yet validated
Scope boundary — no patient uploads into this Workbench or treatment decisions. For a local, nonclinical redacted-document metadata check, use the Evidence Intake. Public catalogue metadata and explicitly public, aggregate research summaries only. No genomic raw reads, patient records, identifiable clinical datasets, treatment decisions, personal predictions or autonomous medical effects. All calculations and exports happen in your browser memory; the selected public catalogue requests go directly from the browser to GDC or cBioPortal. The hosting service may record standard access metadata.
01 · Actual public data sources
Live research catalogues
These read-only APIs show real study/project metadata, not imported tumour profiles and not efficacy evidence. No token, password or patient-identifying query is sent. Availability is subject to the third-party provider's API, CORS and service policies.
Import a locally held CSV of published aggregate statistics. Each row is a source-labelled research summary. A shared cohort label permits coverage checks but never proves that measurements came from matched individuals.
0accepted summary rows
0reported cohorts
0 / 5modalities represented
UNKNOWNPIGA data admission
No dataset loaded. Patient-level identifiers and nonpublic data are not accepted.
Required columns: modality, cohort, feature, n, mean, sd, unit, source_url, access. Modality values: genomics; transcriptomics; proteomics; single_cell_rna; spatial_biology. There is no cross-patient or cross-modality linkage or training from imported rows. Every supplied aggregate statistic remains provenance-dependent.
Enter aggregate results from a documented held-out cohort. Compare a candidate model against a baseline on the same named endpoint, with the same test cohort and data split. This tool computes only a descriptive difference, not statistical significance or superiority.
No benchmark. Independent prediction, holdout leakage control, calibration, bias and external generalizability are not established.
04 · Resistance mechanisms as testable hypotheses
Falsifiable adaptation hypothesis
Document a possible mechanism such as compensatory signalling, selection or microenvironmental adaptation. This is a study-planning record, not a patient-specific resistance forecast.
No mechanism verified. Wet-lab reproducibility and treatment resistance prediction remain unproven.
05 · Actual validation evidence ladder
Wet-lab and preclinical register
Prespecify work for cell lines, 3D organoids and, only where scientifically necessary and ethically authorized, animal models. Recording a plan is not an experiment.
Wet-lab experiments: NOT EXECUTED.
06 · Prospective clinical evidence roadmap
Trial-design readiness (not patient enrollment)
Structure a prospective comparison: standard-of-care versus study-assisted workflow, prespecified outcomes, consent, safety, study oversight and statistical plan. The app neither randomizes people nor issues care advice.
Human trial: NOT INITIATED. Clinical benefit: NOT ESTABLISHED.
07 · Legal and regulatory readiness
SaMD and EU MDR regulatory evidence plan
Software intended for medical diagnosis or treatment support may enter medical-device oversight depending on intended purpose and functionality. Nothing in this page performs classification, CE marking, FDA clearance or approval.
European Union: Regulation (EU) 2017/745 (MDR)
Intended purpose, software qualification/classification, risk management, clinical evaluation, verification, cybersecurity, technical documentation and conformity assessment must be assessed by qualified parties.
Determine whether proposed functions are medical devices and which requirements apply; perform appropriate validation, quality and documentation. No authorization is presumed.
Immutable implementation boundary: research Gate 1 may record ALLOW, DELAY, BLOCK or UNKNOWN. Clinical Gate 2 always returns BLOCK; no form can unlock it. Any hypothetical future medical-purpose software requires a new, separately evaluated and authorized product, never a UI toggle.
08 · Glassbox evidence trail
PIGA / A7SEM / A7SEM Reverse research passport
Seven interacting dimensions are lenses for questioning evidence, not established clinical biomarkers or a time-based stage sequence.
The report distinguishes loaded public metadata, unverified user-supplied aggregates, calculated descriptive contrasts and future study planning. No checklist item confers clinical permission.
From research instrument to evidence-led clinical technology
The five workstreams below are research and regulatory objectives, not capabilities demonstrated by this app. Medical urgency is determined by the patient's condition and a meaningful intervention window; the PIGA action gate controls whether this software is permitted to influence clinical action. An urgent case does not unlock an unvalidated tool.
Clinical action admission: Gate 2 = BLOCK. No patient-specific resistance prediction, treatment-switch recommendation, automated triage, or authorized bedside use. The research app cannot become an approved medical device by checking boxes.
Workstream 1 · Regulatory
Medical-purpose authorization, not a toggle
Needed: a precise intended purpose, EU MDR qualification and Rule 11 risk classification, verification, clinical evaluation, safety, quality management, cybersecurity, and applicable conformity or FDA processes.
MDR Class IIa, IIb or III is risk- and purpose-dependent, not invariably Class III.
FDA Breakthrough Device designation is not marketing authorization; the US pathway depends on the particular function.
Separate medical-purpose product and independently verified permissions are necessary before any authorized clinical deployment.
Needed: prespecified disease-specific use case, standard-of-care comparator, clinically meaningful endpoint and statistically sound analysis; evidence requirements must match claim and regulatory pathway. A demonstrated PFS difference is not an OS benefit.
Prospective study design, approvals, safety monitoring and sufficient follow-up.
External reproducibility, bias controls, subgroup safety and full event adjudication.
Never assert a Phase III trial was performed by PIGA.
Needed: validated sampling and assay intervals, limits of detection, paired molecular/radiological outcomes, confounders, uncertainty bounds and time-to-detection analysis.
Benchmark sensitivity, specificity, PPV, false-alarm rate, calibration and lead time.
Compare any earlier warning to independent, prespecified standard care.
No fabricated clock-time certainty, such as “resistance in exactly 21 days”.
Status: research hypothesis, no individual resistance forecasting.
Workstream 4 · Synthetic lethality
Testable biological vulnerabilities
Needed: a mechanistic network hypothesis (for example, context-specific TP53/KRAS pathway dependencies), matched genetic controls, rescue experiments and effects in independent organoid/cell systems.
Separate tumour-selective activity from general cytotoxicity.
Document off-target effects, normal-cell comparators and independent replication.
In-vitro success is not a clinically validated treatment combination.
Status: discovery design only; no new treatment demonstrated.
Workstream 5 · Clinical guidelines
NCCN, ESMO and S3 evidence translation
Needed: independently replicated patient-relevant benefit, implementation safety, external guideline appraisal and updates to the clinical evidence base.
A professional society guideline is not automatically a binding law or a mandate to purchase named software.
Decisions about standard care depend on context, jurisdiction and clinical judgment.
No guideline endorsement, mandatory tumour-board use or malpractice determination is claimed.
Status: no inclusion or endorsement asserted.
External benchmark — SERENA-6 (2025), not a PIGA result
In a specific ER-positive/HER2-negative advanced breast-cancer population with emerging ESR1 mutations, 315 participants were randomized. At the published 2025 interim analysis, median progression-free survival was 16.0 months with the studied early therapy switch versus 9.2 months with the comparator; this is a PFS finding, not proof of longer overall survival, and neither arm used this workbench. Assay monitoring was study-specific: a proposed two-week cadence or precisely timed individual resistance event is not established here.
Each step requires documented independent evidence and a fresh admission decision. Stages are not automatically earned by completing a form. No endless execution loop and no new PIGA engine are introduced.
AReproducible researchPublic data provenance, held-out comparators, reproducible metrics and uncertainty. Present app: supporting tooling only; independent superiority unverified.
BMechanistic relevancePredeclared biological falsifiers and external cohorts. No verified disease-specific predictor.
CSilent-mode studyEthically reviewed prospective observation; no change to treatment. No study underway.
ERegulated routine useAppropriate authorization and postmarket oversight for a separate defined medical function. Not authorized.
Five requirement specifications (planning only)
Tick a line only if its research requirement has been written down. These are not attestations that evidence, approvals or clinical milestones exist.
0/5 research requirements defined · 0 clinical milestones verified · Gate 2: BLOCK.
Primary independent benchmark: early resistance detection
A falsifiable protocol should state tumour subtype, alteration, clinically relevant decision horizon, sampling intervals, assay limitations, prespecified comparator, time-to-event definition, and independent external validation.
Evaluate more than accuracy
Prespecify sensitivity, specificity, positive predictive value (prevalence-dependent), calibration/Brier score, false alarms per tested patient, time-dependent discrimination, detection lead time, missingness, assay error, subgroup performance and downstream harms.
A7SEM Reverse: compulsory disconfirmation
Before any clinical claim, test spurious signals, data leakage, batch effects, selection bias, lead-time bias, ineffective treatment switches, false positives, population drift and external replication failure. A theoretical network mechanism alone is never clinical evidence.
Fixed clinical admission result: BLOCK. Gate 1 may evaluate research source admissibility. No local checkbox, JSON passport, lab protocol or user-supplied metric can unlock clinical Gate 2. Any later product would require independently documented authorization for a specified indication.