Precision Accelerated Repair Biology: A Bounded Mechanistic Evidence Map for a 20-Material Case Set

Evidence boundary

Preprint. Not peer reviewed. This signed receipt verifies release provenance and integrity, not scientific validity, efficacy, safety, or suitability for human use.

Public contributors

1. Kamil Khoury, author

Abstract

Multicomponent repair proposals routinely join materials with unequal chemical definition, target knowledge, exposure data, and evidentiary maturity. This paper audits a closed historical case set of two physiologically regulated ion contexts and eighteen peptide or peptide-derived labels. Revision 4 retains the multilingual primary-source recovery pass and adds one separately bounded 2026 KPV study in oleic-acid-treated HepG2 cells. It maps bounded propositions concerning mitochondrial state, apoptotic commitment, inflammatory signaling, cell migration, endothelial behavior, matrix organization, neural state, and cell-state persistence. The paper does not claim that the twenty labels form a validated therapeutic system, share pharmacology, or improve a general repair outcome.

The title terms are used only operationally. Precision means selecting a defined biological state by measured constraints, not assigning a product to a tissue label. Acceleration means a prespecified improvement in a recovery-trajectory estimand, such as time to a functional threshold or recovery slope, relative to a matched comparator. Repair means restoration of a named tissue-relevant function together with noninferior prespecified safety states. No such acceleration or repair result is reported here. The contribution is a claim-level evidence map, a transparent identity and uncertainty record, and a narrow synthetic design example. The evidence map is curated rather than systematic; it has an explicit cutoff and does not establish clinical safety, efficacy, compatibility, optimal timing, or benefit of any combination.

**Keywords:** causal inference; systems pharmacology; peptide biology; combination experiments; target engagement; repair biology; identifiability; preclinical research

1. Category definition, case-set provenance, and evidence boundary

“Repair” is not a molecular variable and no molecule is intrinsically pro-repair across tissues, injuries, doses, and times. Viable-cell preservation, inflammatory resolution, endothelial behavior, matrix organization, epithelial barrier restoration, axonal integrity, and durable tissue function are different outcomes with different failure modes. This paper therefore begins with a causal question rather than an aggregate promise.

The 20-label set is inherited from the previous public PARB release and is retained as a closed audit case set, not selected as a representative sample of repair biology or as a recommended portfolio. It contains magnesium, zinc, and eighteen peptide or peptide-derived labels. The number twenty is a property of this historical list, not a biological principle. The list includes physiologically regulated ions, sequence-defined peptides, a copper complex, and a heterogeneous peptide mixture; they cannot be modeled as interchangeable binary agents.

This paper makes no claim that the listed materials:

The paper is useful only if a reader can identify an experiment whose result would remove, reverse, or quantify a proposed causal edge. Revision 4 therefore distinguishes direct cell-count and tissue-outgrowth observations from Ki-67, PCNA, p16/p21, transcript, protein, or in-silico results, rather than collapsing all of them into “regeneration.” It is not a substitute for material analysis, pharmacology, toxicology, controlled experiments, regulatory review, or clinical development.

1.1 The estimand precedes any network

Every proposed experiment must state an estimand in the following form:

In **[defined biological system]** exposed to **[defined injury/stressor]**, what is the effect of **[defined active-species concentration-time perturbation]** on **[prespecified functional outcome]** over **[time interval]**, relative to **[comparator]**, and is that effect consistent with **[prespecified proximal mechanism]** without failure of **[safety/quality guardrails]**?

For an acceleration claim, add a named temporal quantity: time to a prespecified functional threshold, slope of functional recovery over a declared interval, or area under a prespecified recovery curve. For a precision claim, add the measured state variable used to select or stratify the biological system. An endothelial barrier-recovery experiment and a neural mitochondrial-injury experiment require separate graphs, state variables, and outcomes. They may share an evidence database, but they cannot share one fitted universal repair parameter.

2. Evidence-map methods and claim architecture

2.1 Multilingual primary-source recovery map, not a systematic review

This manuscript is a curated evidence map informed by primary experimental reports, primary clinical/regulatory records where relevant, and methodological literature. Revision 4 retains a documented multilingual evidence-recovery pass for every case-set label: English and Russian names, transliterations, literature sequences, product codes, and mechanism/endpoint terms were searched independently. The recovery record is released as `PARB_R3_evidence_recovery_log.md`; the literal-search and extraction protocol is released as `PARB_R3_global_evidence_recovery_protocol.md`. Revision 4 adds a separately captured KPV record from indexed bibliographic metadata and abstract text because the publisher full text was subscription-restricted at review. It is not a systematic review, has not produced a PRISMA flow diagram, and does not estimate pooled efficacy.

The map deliberately does not infer that absence from the registry means absence of evidence. A component-specific conclusion may be upgraded only after the source, exact material identity, experimental system, endpoint, and permitted wording have been extracted. Review articles, product catalogues, patents, theses, conference reports, and narrative papers may identify sources or establish a limited provenance point, but they may not silently provide a causal mechanism. The evidence cutoff is 2026-07-24. The claim-level extraction is released as `PARB_R4_master_evidence_registry.csv` and is supplemented by the Revision 3 recovery log.

Reviews can provide orientation but do not establish causal edges. A paper reporting a marker association is not evidence of direct binding; direct cell counts, EdU/BrdU, clonogenicity, or a validated tissue-outgrowth endpoint must not be restated as one another. A knockout, blockade, rescue, or target-engagement result may support a causal edge in the tested system, but it does not automatically transfer to another cell type, tissue, formulation, or exposure regime.

2.2 Edge-evidence record

Each claim-level relation in the evidence map is a structured record rather than a single favorable citation:

e=(s,t,\operatorname{sign},m,\mathcal S,\tau,\delta,q,\mathcal C),

where `s` and `t` are source and target nodes; `sign` is positive, negative, mixed, or unresolved *for a named measurement*; `m` is the mechanistic proposition; `S` is the exact experimental system; `tau` is the observation window; `delta` is the reported effect metric; `q` is the evidence-quality vector; and `C` is the complete citation and material record. The quality vector is

q=(I,C,R,D,X,P,U),

with identity confidence (`I`), causal support (`C`), independent replication (`R`), biological directness (`D`), exposure plausibility (`X`), precision/risk of bias (`P`), and residual uncertainty (`U`). These dimensions must be displayed, not silently collapsed into a numeric “evidence weight.” Evidence strength is not effect magnitude.

For editorial readability, a claim can additionally receive one of four labels:

No label applies to an entire component. It applies to one proposition in one biological context. In this paper, “Observed” means a result reported by the cited study, not a result observed by the author. “Causal support” requires a material-verified perturbation plus at least one dependency test or orthogonal perturbation in the named system; it is not a claim of universal mechanism.

The registry records each quality field as high, medium, low, or not assessed and reports residual uncertainty separately. It does not collapse those fields into a score. The present revision contains only claims for which a source, system, and permitted wording can be recorded; missing fields remain explicit rather than inferred.

3. Molecular inventory, active species, and material identity

3.1 Three levels that must not be collapsed

Every candidate has three distinct representations:

Only a verified biological perturbation belongs in a causal-effect model. Material and active-species information belong in the exposure and measurement layers. This distinction prevents a proprietary label from becoming a fictional receptor ligand.

3.2 Reconciled case-set inventory

The following table is a working nomenclature table, not a certificate of material identity. A literature sequence still requires verification of stereochemistry, terminal chemistry, salt/counterion, purity, lot, and stability in the material actually tested. “Unresolved” is not a negative efficacy conclusion; it prohibits receptor, occupancy, pharmacokinetic, or interaction parameters before the relevant data exist. Claim-level source and identity status are in the registry.

Earlier versions of this paper mistakenly equated Livagen with Vilon and Pancragen with Cartalax. Those are identity errors, not minor nomenclature issues. They invalidate duplicate-node arguments, molecular-weight tables, pair counts, and all downstream interaction claims. The 20-label list must never be described as 20 chemically distinct active species unless a material-release record demonstrates that fact.

3.3 Minimum material-release record

Before any experiment, record sequence, stereochemistry, terminal modifications, salt/counterion, average and monoisotopic mass, formula where available, complex stoichiometry, purity method and result, impurities/degradants, concentration assay, lot, solvent, pH, ionic strength, container, storage, and stability in the actual biological matrix. Cerebrolysin additionally requires a lot-level peptide fingerprint and a potency assay. Mineral experiments require free and total ion measurements where feasible, relevant ligands, and compartmental context.

4. Granular molecular pharmacodynamics and pathway architecture

The following modules summarize only the claims represented in the registry. They are not exhaustive reviews. For each claim, the reader should consult the registry for the tested system, source class, correction status, uncertainty, and permitted wording. A literature observation is never converted here into a product-specific pharmacokinetic, tissue-homing, safety, or efficacy claim.

4.1 Mitochondrial survival module

Elamipretide has cardiolipin-associated mitochondrial literature. The 2014 citation is a review and is used only for orientation; it is not treated as a causal edge. Testable proximal measurements include cardiolipin oxidation or organization, membrane potential, respiratory organization, coupling, oxidant emission, and cytochrome-c behavior. ATP-linked respiration alone does not validate a complete causal chain. Elamipretide should not be described as a conventional surface-receptor agonist. A current translational fact must remain separate from this mechanistic hypothesis: FDA granted accelerated approval to Forzinity (elamipretide) on 2025-09-19 for improving muscle strength in qualifying patients with Barth syndrome. That narrow approval does not validate any PARB combination, route, dose, disease claim, or mechanism outside the label. [Szeto, 2014](https://pubmed.ncbi.nlm.nih.gov/24117165/); [FDA, 2025](https://www.fda.gov/news-events/press-announcements/fda-grants-accelerated-approval-first-treatment-barth-syndrome)

MOTS-c has reported metabolic effects involving folate/purine perturbation, AICAR accumulation, AMPK-associated signaling, and stress-dependent nuclear translocation. AMPK is not a unidirectional repair signal: transient adaptive signaling can differ from persistent energy-conservation signaling that restrains biosynthesis. A valid model therefore measures metabolite context, AMPK kinetics, localization where feasible, and functional recovery separately. [Lee et al., 2015](https://pubmed.ncbi.nlm.nih.gov/25738459/); [Kim et al., 2018](https://pubmed.ncbi.nlm.nih.gov/29983246/)

Humanin has direct experimental support for interference with BAX activation in selected systems. That supports a conditional edge to the intrinsic-apoptosis decision process, not correction of upstream injury. All Humanin experiments must pair short-term viability with mitochondrial integrity, DNA-damage response, senescence, abnormal persistence, and durable function. Receptor-associated pathways are separate hypotheses and cannot be collapsed into BAX interference. [Guo et al., 2003](https://pubmed.ncbi.nlm.nih.gov/12732850/)

Magnesium enters this module through ion activity, MgATP chemistry, and enzyme kinetics; it does not add energy as a universal scalar. Revision 4 contains no extracted component-specific magnesium edge and therefore makes no directional biological claim. A measured magnesium-limited state is a context condition to be tested, not a benefit inference.

4.2 Inflammatory and epithelial/glial-response module

KPV is distinct from full-length alpha-MSH and from gamma-MSH. In defined intestinal epithelial models, PepT1-mediated transport and reduced NF-kB/MAPK-related inflammatory signaling have been reported. That is a transporter- and tissue-context-dependent result, not proof of systemic melanocortin agonism. Functional tests must include barrier integrity and host-defense/clearance measures, because a lower cytokine signal can be either beneficial resolution or harmful suppression. [Dalmasso et al., 2008](https://pubmed.ncbi.nlm.nih.gov/18061177/); [Land et al., 2012](https://pubmed.ncbi.nlm.nih.gov/22837805/)

Separately, Lee and colleagues reported an oleic-acid-treated HepG2 cell model in which KPV at 100 micrograms per milliliter was associated with lower lipid accumulation and fatty-acid-synthase expression, without cytotoxicity in that assay. The authors also reported lower reactive-oxygen-species generation, prevention of extracellular-signal-regulated-kinase activation, lower AKT phosphorylation, lower mTORC1 phosphorylation, and altered PPARgamma phosphorylation under that model's steatosis conditions. This is a cell-model observation, not evidence of human hepatic efficacy, safety, dose, pharmacokinetics, systemic activity, direct receptor occupancy, PepT1 dependence, or a general KPV mechanism. The result requires independent replication, exposure-response characterization, material identity confirmation, and direct dependency testing before it can support a broader causal edge. [Lee et al., 2026](https://pubmed.ncbi.nlm.nih.gov/42064835/)

Cibinetide is an erythropoietin-derived 11-mer with non-erythropoietic tissue-protection hypotheses. A rat neuropathic-pain study reported pain and microglial phenotypes, while a separate receptor study did not support an EPO-promoted EPOR–beta-common extracellular association. Neither result establishes a universal glial-state or receptor mechanism. Receptor-component expression, direct binding or dependency, and loss-of-function testing are mandatory in each responsive lineage. The receptor-counterevidence paper has a linked correction that must travel with the citation. [Swartjes et al., 2014](https://pubmed.ncbi.nlm.nih.gov/24529189/); [Cheung et al., 2018](https://pubmed.ncbi.nlm.nih.gov/30127368/); [correction](https://pubmed.ncbi.nlm.nih.gov/31110193/)

4.3 Endothelial, migration, and matrix module

BPC-157 has been associated with endothelial migration/tube phenotypes and increased VEGFR2 expression/internalization with VEGFR2–Akt–eNOS signaling in defined endothelial and rodent ischemia studies. This supports a candidate mediated path, not direct VEGFR2 ligand identity. Causal tests must distinguish receptor abundance, trafficking, phosphorylation, Akt/eNOS state, nitric-oxide environment, migration, barrier integrity, and mature vessel organization. [Hsieh et al., 2017](https://pubmed.ncbi.nlm.nih.gov/27847966/)

Authentic thymosin beta-4 formed a 1:1 complex with actin monomers and inhibited polymerization in a primary biochemical study. Tissue-level consequences remain context dependent: a closure or speed metric alone is not repair. Its relation to BPC-157 is a hypothesis requiring a factorial study, not established synergy. [Safer, Elzinga, and Nachmias, 1991](https://pubmed.ncbi.nlm.nih.gov/1999398/)

GHK-Cu must remain a coordination-chemistry problem. GHK, copper-only, and Cu(II)-GHK are distinct conditions. A structural study found that Cu(II)-GHK solution chemistry differs from solid-state organization and involves defined ligand coordination; that supports the need for actual-species measurement, not a universal biological mechanism. The often repeated broad transcriptional signature is review-derived and is not evidence of a direct master receptor. Any matrix experiment must compare vehicle, GHK, matched copper-only, and analytically defined Cu-GHK across early chemistry, protein/activity, structure, function, and fibrosis guardrails. [Hureau et al., 2011](https://pubmed.ncbi.nlm.nih.gov/21780203/); [Pickart and Margolina, 2018](https://doi.org/10.3390/ijms19071987)

Zinc is treated here as a metalloprotein, transporter, redox, and transcriptional context variable. Revision 4 contains no extracted component-specific zinc edge and makes no directional biological claim. It can be investigated only after free/bound zinc, transport context, and a named enzyme or barrier process are measured.

4.4 Telomere and cell-state module

Epitalon has reports of telomerase activation and telomere changes in cell culture. These are not organismal rejuvenation evidence. Telomerase activity, telomere-length distribution, senescence, cell-cycle state, genomic stability, differentiation, and long-term function are separate endpoints. A 2025 cell-line study reporting telomere extension with telomerase or ALT-associated findings makes tumor-state guardrails a direct requirement. The paper has a linked erratum, which must be checked before interpreting the result. [Khavinson, Bondarev, and Butyugov, 2003](https://pubmed.ncbi.nlm.nih.gov/12937682/); [Al-Dulaimi et al., 2025](https://pubmed.ncbi.nlm.nih.gov/40908429/); [correction](https://pubmed.ncbi.nlm.nih.gov/41240216/)

4.5 Mixture and recovered short-peptide modules

Cerebrolysin is a mixture. It has no single molecular weight, receptor-occupancy curve, half-life, or interaction coefficient. The cited analytical study profiled an Internet-obtained sample and identified many peptide species; it does not establish product equivalence, an active fraction, or a functional mechanism. Reported neurotrophic-system or microglial phenotypes remain mixture-level observations until lot composition, concentration, target, and causal dependency are resolved. [Gevaert et al., 2015](https://pubmed.ncbi.nlm.nih.gov/26017115/)

Revision 4 retains the prior replacement of blanket “unresolved” treatment for Pinealon, Cortagen, Livagen, Vilon, Pancragen, Cartalax, Chonluten, and Vesugen with source-bounded observations. This is an evidence recovery, not a relaxation of the causal standard. The recovered literature includes direct counted human adipose-MSC increases with H-Lys-Glu-Asp-OH in a 2013 conference report; KED-associated p16/p21 reduction in late-passage oral MSC cultures; KED-, EDR-, and related-peptide tissue-outgrowth results in rat organotypic systems; Pinealon-associated oxidative-stress, propidium-iodide-defined necrotic-death, ERK1/2, and cell-cycle observations; defined AED, KE, KEDA, KEDW, and AEDP assay findings; and an explicit source-internal Chonluten/Vesugen label conflict. The relevant results are preserved in Section 8 and Appendix A with their assay classes and limitations.

These observations do **not** establish a common receptor, direct promoter occupancy, tissue selectivity, bioavailability, pharmacokinetics, formulation equivalence, safe combination, or human benefit. A modelled peptide-DNA interaction is an in-silico hypothesis unless confirmed by orthogonal binding and dependency work. Ki-67, PCNA, p16/p21, p53, or transcript/protein changes are surrogate or state measurements; they cannot be rewritten as direct cell expansion. Conversely, the Vesugen human adipose-MSC report used serial cell counts and therefore must not be downgraded to a marker-only result.

For every recovered short peptide, the advancement sequence is: verify material, termini, stereochemistry, purity, stability, and metabolites; reproduce the exact reported phenotype in the stated lineage; separate cell number from migration and outgrowth; establish a concentration-time relation; identify an early proximal perturbation; test its necessity with a dependency or orthogonal perturbation; replicate outside the originating source network; and connect the result to a lineage-relevant function with cancer, fibrosis, immune, genomic-stability, and abnormal-survival guardrails. Existing ex-vivo, cell-line, animal, primate, thesis, or conference observations can nominate this sequence but cannot supply a universal interaction coefficient.

4.6 Cross-mechanism integration: what may be complementary, conflicting, or unresolvable

The proposed architecture is not a twenty-way receptor cocktail. Its intelligible form is a set of candidate layers that may become complementary only in a defined injury state. Elamipretide, MOTS-c, Humanin, and measured magnesium status occupy distinguishable portions of a mitochondrial-survival question: membrane organization, metabolic adaptation, apoptotic commitment, and ATP-dependent chemical context. The predicted relation is not that every input increases “energy.” It is that an early membrane-state perturbation, an adaptive metabolic signal, and an apoptosis-threshold perturbation could be measured as different links in one stress-to-function trajectory. Their effects can be neutral, redundant, antagonistic, or favorable depending on the stressor and phase.

KPV and Cibinetide illustrate a second potential layer: both have inflammation-related observations, but they belong to different cellular contexts and do not justify a shared systemic anti-inflammatory coefficient. BPC-157 and thymosin beta-4 illustrate signaling-plus-execution complementarity: a candidate endothelial state perturbation and an actin-regulatory state might both be necessary for organized movement, while neither proves durable vascular function alone. GHK-Cu, zinc, and matrix biology present an explicitly bidirectional compatibility problem because redox state, ion speciation, protease balance, stiffness, and fibrosis can move in opposite directions.

The neural-state candidates are the least resolved group. Cerebrolysin is a lot-defined mixture; Pinealon and Cortagen are short peptides with unresolved direct targets; Cibinetide, Humanin, MOTS-c, and Elamipretide have distinct injury-response or stress hypotheses. Their convergence on a neural endpoint must never be rewritten as one neuroregenerative pathway. The appropriate integration is a measurement plan across neuronal survival, glial state, axonal or synaptic architecture, electrophysiology, vascular coupling, and durable function.

4.7 Translational and regulatory boundary as of the evidence cutoff

Regulatory status is not evidence of a general mechanism, but it is relevant to publication accuracy and safety interpretation. As of the cutoff, elamipretide has a narrow FDA accelerated-approval indication for qualifying patients with Barth syndrome; no inference about the other case-set labels follows. FDA also identifies important safety-information and material-characterization gaps for compounded BPC-157 and MOTS-c related substances. The FDA public calendar and briefing-material page records a July 23 to 24, 2026 Pharmacy Compounding Advisory Committee discussion of BPC-157-, KPV-, TB-500-, and MOTS-c-related bulk substances. This revision does not claim a committee recommendation or final FDA action because no such determination was verified from that source at the evidence cutoff. [FDA Forzinity record](https://www.fda.gov/drugs/drug-trials-snapshots/drug-trials-snapshots-forzinity); [FDA bulk-substance safety record](https://www.fda.gov/drugs/human-drug-compounding/certain-bulk-drug-substances-use-compounding-may-present-significant-safety-risks); [FDA advisory-committee calendar](https://www.fda.gov/advisory-committees/advisory-committee-calendar/july-23-24-2026-meeting-pharmacy-compounding-advisory-committee-07232026)

5. Candidate-relation map and physical-compatibility boundaries

The registry is the claim-level relation map for this revision. It is intentionally not a complete interactome: most label pairs have no verified material identity, exposure data, or single-agent causal edge. The pair table in Section 4.6 lists hypotheses and conflicts, not established arrows. No physical compatibility is inferred from shared therapeutic intent.

The relevant experimental material can change with pH, ionic strength, buffer, co-solvents, protein binding, container surface, temperature, freeze–thaw history, oxidation, adsorption, aggregation, metal coordination, and time. A common mixture with unknown integrity turns chemical incompatibility into an unmeasured confounder of every biological inference.

For sequence-defined peptides, the minimum compatibility assessment is intact-material quantification in the actual exposure matrix, degradation-product analysis, aggregation/particle assessment where applicable, pH and osmolality, and retained biological activity. For GHK-Cu, copper speciation, oxidation state, competing ligands, and peptide integrity are essential. For magnesium and zinc, total concentration is insufficient: free-ion activity and relevant binding/transport context determine the biological perturbation. Cerebrolysin requires lot-specific analytical characterization and an empirical potency assay.

The evidence map consequently represents uncertain materials as separate inputs, u_i(t), rather than as a fictional stable common solution. Separation is not an administration recommendation; it is an experimental design choice that preserves attribution. No generic diluent, preservative, storage interval, or beyond-use interval is established for the case set.

6. Quantitative systems model and non-linear dynamics

6.1 General principles

An evidence graph is not a fitted dynamical model. The graph stores propositions and their evidence; an executable model is fitted only in one tissue/injury context with defined measurements. The model must distinguish: material input; active-species concentration; target engagement or physical interaction; latent or measured state; assay observation; functional outcome; and adverse state.

All equations below are templates. They become numerical only when variable definitions, units, initial conditions, parameter estimates or priors, sampling times, observation error, and identifiability analyses are supplied. If those requirements are not met, the correct output is a qualitative evidence-annotated graph, not simulated efficacy curves.

6.2 Exposure and active-species layer

Let C_i(t) denote the measured concentration of an analytically defined active species in a declared compartment. For a sequence-defined material with suitable data, a minimal parent-species observation model is

\frac{dC_i}{dt}=R_{in,i}(t)-k_{obs,i}C_i(t),

where R_in,i has concentration/time units and k_obs,i has inverse-time units. If parent concentration alone is observed, elimination and degradation are ordinarily identifiable only through their sum. Separate elimination and degradation rates require metabolite, mass-balance, or independently informative data. The model is not valid for a mixture unless C_i identifies a particular fraction, and it is not an ion-homeostasis model for magnesium or zinc.

For a verified 1:1 equilibrium target and measured free ligand concentration L_f at the target-relevant compartment, fractional occupancy can be represented by

O_i(t)=\frac{L_{f,i}(t)}{K_{d,i}+L_{f,i}(t)}, \qquad 0\le O_i\le1.

K_d is a binding parameter under stated equilibrium assumptions; it is not interchangeable with an EC50. When the intended quantity is an empirical response rather than occupancy, use a separately labelled Emax/Hill response function and estimate EC50 and Hill slope in the selected system. A pathway marker, tissue label, or favorable phenotype is not occupancy. When a target is unknown, exclude occupancy from numerical causal calculations and use measured exposure with an explicitly empirical response model.

6.3 Magnesium example: energy chemistry rather than an additive gain

The energetic state should not merge ATP ratios, ATP-linked respiration, and phosphocreatine into one variable. A more chemically interpretable starting point is

A_T=[ATP]+[ADP]+[AMP], \qquad
EC=\frac{[ATP]+\tfrac12[ADP]}{A_T},

\frac{d[ATP]}{dt}=J_{OXPHOS}+J_{SLP}-J_{ATPase}-J_{transfer}.

For a single 1:1 MgATP equilibrium, total ATP and total magnesium cannot be substituted for free ATP and free magnesium in a mass-action expression. Let x be the MgATP concentration, A_tot total ATP, M_tot total magnesium, and K_d the conditional dissociation constant in the declared matrix. Conservation and equilibrium give

x=\frac{(A_{tot}+M_{tot}+K_d)-\sqrt{(A_{tot}+M_{tot}+K_d)^2-4A_{tot}M_{tot}}}{2}.

This is still only a conditional single-ligand approximation. Numerical inference requires ionic strength, protonation, competing ligands, compartments, and a decision about free versus total analytical measurements. Magnesium can then enter a named measured enzyme-rate law; it must not be represented as a universal positive ATP flux.

6.4 Phase-specific biological-state model

Let `X_mn(t)` be a measured state `m` in biological replicate `n`; choose a reference mean and scale before fitting and define

\xi_{mn}(t)=\frac{X_{mn}(t)-\mu_{m,ref}}{s_{m,ref}}.

In a defined phase `p(t)`, a sparse stochastic model is

d\boldsymbol\xi_n(t)=
\left[A_{p(t)}\boldsymbol\xi_n(t)+B_{p(t)}\mathbf E_n(t)
+\sum_{(i,j)\in\mathcal H}\mathbf c_{ij,p(t)}E_{in}(t)E_{jn}(t)
+\mathbf b_n\right]dt+L_{p(t)}dW_n(t).

`A`, `B`, and `c` have inverse-time units when `xi` and `E` are dimensionless. The replicate term must be assigned a distribution and units, for example b_n distributed as Normal(0,Sigma_b) with inverse-time units; a constant replicate drift can otherwise create nonstationary divergence. The interaction set `H` must be small, preregistered, and justified by single-agent mechanism data. It is not the complete set of 190 product-label pairs. Bounded states require a link or transformation, and the measurement matrix requires a stated scale and error distribution.

Assays observe states imperfectly:

\mathbf y_n(t_k)=H_{obs}\boldsymbol\xi_n(t_k)+\boldsymbol\epsilon_{nk},

with assay-specific error distributions and a declared covariance structure. A latent state should not be introduced merely to relabel a collection of unrelated assays. It requires a justified measurement model.

6.5 Functional outcomes and non-compensatory safety

The primary analysis should target a direct functional measurement, not a universal repair score. For example,

h\{\mathbb E(Y_{nk}\mid \mathbf u_n)\}=\alpha+\boldsymbol\gamma^T\boldsymbol\xi_n(t_k)+\boldsymbol\eta^T\mathbf u_n+u_{batch}+u_n,

where `h` and the full outcome distribution are appropriate to the endpoint, such as a binomial or beta-binomial model for a fraction or a count distribution for a positive count. The treatment vector remains explicit when the estimand is a total treatment effect; conditioning on post-treatment states alone does not identify total effects or formal mediation. The functional endpoint must be specific to the context: barrier resistance, controlled closure, directed migration with intact architecture, perfusion-related function, electrophysiology, or another declared outcome.

If a secondary composite is required, predefine bounded desirability functions `d_l(Y_l)` and use safety gates rather than allowing harm to be numerically offset by benefit:

R_{secondary}=\left(\prod_{l=1}^{L}d_l(Y_l)^{w_l}\right)
\times\left(\prod_{k=1}^{K}\mathbb 1[G_k\ \text{passes}]\right),
\qquad \sum_{l=1}^{L}w_l=1.

Here `I(condition)` is an indicator: it equals 1 when its condition holds and 0 otherwise. Thus any failed prespecified guardrail sets the secondary composite to zero rather than allowing a favorable desirability term to offset the failure.

Guardrails can include genomic-damage persistence, abnormal survival/proliferation, impaired host defense, barrier leak, disorganized or fibrotic matrix, excessive angiogenic behavior, or decline in tissue-specific function. A failed guardrail invalidates a favorable composite interpretation.

6.6 Combination effects and the meaning of synergy

For binary inputs at one fixed dose, the contrast

\Delta_{ij}=\mu_{11}-\mu_{10}-\mu_{01}+\mu_{00}, \qquad \mu_{ab}=\mathbb E(Y\mid u_i=a,u_j=b)

is an additive-scale interaction contrast. It is not a general biological synergy coefficient. It changes under nonlinear scales, ceilings, floors, and different outcome directions.

For a selected pair, obtain replicated concentration-by-concentration and time-resolved data, then fit an interaction surface such as

h\{\mathbb E[Y_n(d_i,d_j,t)]\}=
\alpha+f_i(d_i,t)+f_j(d_j,t)+f_{ij}(d_i,d_j,t)+u_{batch}+u_n.

The functions must be constrained so that f_i(0,t)=f_j(0,t)=0 and f_ij(d_i,0,t)=f_ij(0,d_j,t)=0; otherwise main and interaction functions are not separately identifiable. The null model must be selected before data analysis. Bliss independence describes a specific probabilistic noninteraction convention for a normalized bounded fractional effect, E_B=E_i+E_j-E_iE_j; Loewe additivity requires a defensible dose-equivalence assumption; highest-single-agent describes a practical but weaker comparator. Parameterized response-surface approaches such as MuSyC can distinguish potency from efficacy interactions. [Meyer et al., 2019](https://pmc.ncbi.nlm.nih.gov/articles/PMC6675406/); [Ianevski et al., 2020](https://academic.oup.com/nar/article/48/W1/W488/5815821)

Call a result synergistic only if the prespecified null is exceeded in a reproducible dose/time region with uncertainty intervals, biological replication, and functional, not marker-only, confirmation. A plausible division of labor between signaling and cytoskeletal execution is **mechanistic complementarity**, not demonstrated synergy.

6.6a Theoretical mechanistic-complementarity framework and falsifiable predictions

The absence of a proven interaction coefficient does not prevent a paper from stating a **pre-registered theoretical prediction**. It does require the prediction to be directional only at the level justified by the single-agent evidence and to name the result that would falsify it. In this paper, a theoretical interaction is a proposed division of labor between measured states; it is never an assertion that all listed materials, doses, routes, or time windows work together. Timing refers to the order of measured biological states in one model, not an administration schedule.

For a candidate pair `i,j`, let `T_ij` denote the prespecified temporal-order criterion, `D_ij` the pathway-dependency criterion, `F_ij` the functional interaction criterion under the chosen null, and `G_ij` the full safety/quality-gate criterion. The following is a promotion rule for interpretation, not a biological effect-size equation:

S_{ij}^{support}=I(T_{ij})\times I(D_{ij})\times I(F_{ij})\times I(G_{ij}).

Here `I(condition)` equals 1 only when that criterion is met. A result can be described as support for a mechanistic interaction hypothesis only when all four terms equal 1; otherwise it remains a single-agent result, a negative result, or an unresolved interaction. In particular, a favorable terminal cell count does not compensate for absent temporal order, an untested pathway, failed function, persistent DNA damage, abnormal survival, impaired host defense, disorganized matrix, or another failed guardrail.

The most useful theoretical prediction is therefore a **state sequence**: an early candidate proximal perturbation must occur before a downstream state change, which must precede the declared function, while adverse-state measurements remain acceptable. If the functional result occurs without the proposed early change, if blocking the proposed state does not attenuate the result, or if the order reverses, the named mechanistic hypothesis is rejected even if an empirical interaction remains possible by another path.

The table deliberately contains favorable, null, antagonistic, and exclusion predictions. That is what makes it adversarially useful. A proposed interaction is not strengthened by a long list of components; it is strengthened when the experiment can distinguish its predicted trajectory from redundancy, independent additivity, pathway substitution, direct stress relief, altered cell composition, assay artifact, or harm.

6.6b Interaction classes and expected empirical patterns

Before a factorial experiment, each nominated pair must be assigned one of the following mutually exclusive theoretical classes. The class is a pre-analysis hypothesis and may be rejected by the data.

6.7 Identifiability, calibration, and validation

Twenty product labels, multiple latent states, and all pairwise interactions contain far more free parameters than a practical initial data set can estimate. The first executable model should contain only a few active species with measurable proximal readouts. Fit models with shrinkage or hierarchical priors, conduct structural and practical identifiability analyses, inspect prior and posterior predictive behavior, compare to simpler alternatives, and validate on an independent batch or model. Time-resolved data are needed to distinguish altered injury onset from altered recovery slope; terminal measurements cannot do so reliably.

7. Biomarker architecture, causal tests, and experimental advancement gates

The evidence map advances individual claims, not whole products or a whole combination.

7.1 Staged research sequence

**Stage 0: analytical release.** Establish identity, purity, stability, active species, concentration assay, and lot comparability. Unknown materials do not advance.

**Stage 1: choose one context.** Define tissue/cell lineages, injury/stressor, injury severity, reference state, primary endpoint, observation window, and veto outcomes. Examples include endothelial/epithelial barrier reconstruction or neural mitochondrial injury; they require separate models.

**Stage 2: single-agent profiling.** Use vehicle, material controls, appropriate positive controls, concentration-time ranges bounded by cytotoxicity and exposure plausibility, biological replication, and batch blocking. Measure active species, proximal biology, intermediate state, function, and delayed adverse state.

**Stage 3: causal deconvolution.** Use genetic perturbation, blockade, rescue, transport inhibition, or orthogonal intervention to distinguish target dependence from a metabolite, nonspecific stress, or altered cell composition.

**Stage 4: pair prioritization.** Advance only pairs whose single-agent windows and mechanistic rationale are established. Use full dose-response surfaces, an interaction null selected a priori, multiplicity control, and at least two independent biological batches. Biological batches are not external replication; confirmatory work should distinguish both.

**Stage 5: higher-order combinations.** Use sequential model-based selection or a prespecified fractional-factorial screen only after validated pair modules. A fractional design must declare which effects are estimable and which interactions are aliased. Confirm candidates independently; do not begin with a 20-component mixture.

**Stage 6: organotypic and in-vivo confirmation.** Re-establish exposure, target engagement, function, persistence, and guardrails. Randomize, blind where feasible, justify sample size, predefine exclusions, report all observations, and follow ARRIVE 2.0 for animal work. [Percie du Sert et al., 2020](https://pubmed.ncbi.nlm.nih.gov/32663219/)

8. Complete component atlas and evidence-disposition map

This table defines what each candidate can legitimately contribute before a combination claim. It is intentionally conservative: an unresolved target does not become known because a product has a tissue-associated name.

8.1 Prioritization tiers

For a first executable prototype, only inputs with defined material and a measurable proximal readout should be considered. Elamipretide, MOTS-c, Humanin, KPV, and thymosin beta-4 can serve as mechanism-focused prototypes in appropriate contexts; Cibinetide requires explicit treatment of receptor uncertainty. BPC-157 and GHK-Cu are pathway-associated candidates that require active-species and causal-pathway deconvolution before interaction coefficients. The recovered short-peptide findings belong in the **evidence-recovery atlas**, not a combination-ready node set: they have source-bounded phenotype observations but generally lack lot identity, exposure, target dependence, independent replication, and safety conditions. Magnesium and zinc should be modeled as measured physiological contexts unless a specific ion-homeostasis experiment is the subject.

The triage rule is mechanical: a material is eligible for a first executable prototype only if the registry records at least medium identity confidence, a measurable proximal readout, and no unresolved material-release gate. It is conditional when a primary claim exists but exposure or causal dependency is unresolved. It is ineligible when product identity is unresolved or the source is only a descriptive phenotype. This is not a clinical ranking and does not infer benefit from inclusion in a higher tier.

9. Temporal-state architecture and spatial/formulation boundaries

No calendar-based or route-specific sequence is inferred by this evidence map. Timing is a biological-state question. An intervention can be tested only when the state it is proposed to affect is both present and measurable. An anti-apoptotic pressure modifier may be interpretable during acute BAX-mediated injury but inappropriate after normal clearance should occur. A metabolic conservation signal may help early stress adaptation but oppose later biosynthetic reconstruction. Inflammatory attenuation may be beneficial after pathological amplification but harmful if it prevents required clearance. Endothelial migration and matrix deposition require structural organization and maturation measures, not simply early maxima.

The model therefore uses state-defined windows:

These are experimental windows, not exposure instructions. Quantitative timing requires component-specific concentration-time, target-engagement, state-turnover, and interaction data in the chosen model.

9.1 Temporal prediction rules for candidate interaction studies

The state windows above permit a concrete but bounded timing theory. At baseline, establish material identity, active-species exposure, injury-state reproducibility, and the bottleneck that limits function. During acute containment, test only candidate modules predicted to affect measurable membrane, energetic, apoptotic, or pathological inflammatory pressure. During structural reconstruction, test only modules with confirmed early-state effects and measure organization, barrier, directed movement, or matrix mechanics rather than raw growth or speed. During maturation/persistence, test whether the early phenotype remains functional after perturbation removal and whether genomic, fibrosis, immune, and abnormal-survival guardrails remain acceptable.

For every proposed trajectory, the preregistration must state: (1) the earliest observable state expected to change; (2) the latest time by which it must change; (3) the downstream state it is predicted to precede; (4) the functional endpoint that is expected to improve; (5) the adverse state that vetoes the claim; and (6) the counter-hypothesis that would explain the same terminal result. A calendar hour can be used only after the state-transition times are measured in that exact model. This prohibits converting a theoretical phase relation into a clinical timing or dosing instruction.

The first factorial study should therefore be a two-agent, four-cell, time-resolved design in one injury state, not a sequential or simultaneous mixture of the case set. The Elamipretide × Humanin prototype is the appropriate first example because it has a specific early-to-late chain to test. MOTS-c may enter only as a separately qualified third-factor extension after its exposure, metabolite, localization, and phase-specific effects are measured. BPC-157 × thymosin beta-4 is a separate endothelial/cytoskeletal program; it should not be merged with the mitochondrial model. KPV/Cibinetide and metal/matrix hypotheses likewise require their own lineage-specific models and safety gates.

10. Research record, reproducibility, and publication governance

10.1 Minimum reporting package

A submission extending this evidence map should provide:

10.1a Revision-3 disclosure status

This is a pre-release draft. The author has not yet supplied affiliation, contributions, funding, competing-interest, material-source, intellectual-property, archive, or license statements for publication. Those facts must be supplied in the release manifest before the status changes to Released Revision 4. The absence of a disclosure statement is not evidence of absence of a relevant interest.

10.2 Interpretation rules

The following inferences are prohibited:

10.3 Explicit limitations of this theoretical reference

An evidence map cannot solve missing chemistry, absent exposure data, non-replicated biology, or poor experimental design by adding mathematical detail. It can make uncertainty explicit and prevent it from being silently converted into parameters. Its graph is inevitably incomplete and context-bound. Its proposed mechanisms should be treated as competing hypotheses, not a map of established physiology. This paper also does not substitute for regulatory, toxicological, manufacturing, or clinical-development requirements.

11. Integrated interpretation and advancement logic

The useful scientific question is not whether a large named set can be narrated as comprehensive repair biology. It is whether a precisely defined material can alter one measured bottleneck in one injury context, by a causal path that predicts a functional result and survives safety guardrails. That question permits genuine accumulation: an identity correction improves the input layer; an exposure assay improves the pharmacology layer; a dependency experiment improves one edge; a functional replication improves a module; and a failed study removes a route rather than being buried under a composite score.

This approach changes the role of combinatorics. The number of possible binary subsets is not a justification for a broad intervention. It is a reason to reduce the search space before combination experiments. Mechanistic plausibility alone does not reduce it enough; the reduction must be earned through verified material, measured exposure, proximal causality, functional relevance, and reproducibility. A sparse model that predicts a blinded intervention in one well-characterized system is more informative than a dense diagram of favorable arrows.

The case set is valuable precisely because it exposes common category errors: regulated ions treated as receptor drugs; sequence labels treated as identity certificates; mixtures treated as single molecules; pathway association treated as direct binding; broad gene signatures treated as direct regulation; and interaction rhetoric treated as synergy. Correcting those errors does not diminish the research program. It makes every positive finding interpretable and every negative finding useful.

11.1 State-gated systems hypothesis for a multi-material research program

The paper can state an **overall systems hypothesis**, but not an overall demonstrated synergistic effect. Here, “system” means a decision-controlled experimental program in which candidate modules may be admitted one at a time as a measured biological state changes. It does **not** mean a simultaneous or sequential twenty-material regimen, a shared formulation, an administration schedule, or a claim that every material contributes favorably. At the present evidence level, the 20-label case set has no estimable net effect and no established all-material interaction coefficient.

The pre-registered systems hypothesis is a conditional state trajectory: qualified injury and a measured bottleneck → reduced acute pathological pressure → organized reconstruction → durable, tissue-relevant function with acceptable persistence, genomic, fibrosis, barrier, and host-defense states. The proposed whole-program prediction is not that every candidate improves every phase. It is that a given phase may be advanced only by a material or pair that has passed the prior phase’s identity, exposure, causal, functional, and safety gates in that exact model. A null, redundant, antagonistic, or unsafe module is removed from that branch; it is not averaged into a favorable composite or replaced by a loosely similar label.

For a specified candidate program with qualified baseline `Q_0` and tested modules `m=1,...,M`, let `P_m` denote the required state-order/dependency result, `F_m` the prespecified functional result, and `G_m` the full guardrail result for module `m`. The following is an admission rule for a claimed system trajectory, not an efficacy equation:

A_{\mathrm{system}}=I(Q_0)\times\prod_{m=1}^{M}\left[I(P_m)\times I(F_m)\times I(G_m)\right].

The system-admission indicator equals 1 only when every predeclared branch in that **particular** experimental program passed its gates. It does not quantify benefit, establish a clinical effect, transfer to another injury context, or validate an untested material. If any term is zero, the system-level interpretation fails for that trajectory and the report must identify the failed module, state, and guardrail rather than report a net “repair score.”

The appropriate empirical test of this architecture is therefore not a single factorial screen of all materials. It is a sequence of small, separately powered, time-resolved studies with locked decision rules. A pair can enter only after its members pass single-agent gates; a later module can enter only after the earlier state transition is observed; and a higher-order model can enter only after lower-order interactions are identified and replicated. The system-level comparator must be predeclared, for example, the best qualified earlier-stage program or a matched minimal module set, and every module-level outcome, failure, and exclusion must remain visible.

This creates a legitimate answer to the question of “synergy over time”: the expected scientific result is a **conditional, state-gated trajectory** that may exhibit sequential complementarity in one defined model. It becomes evidence of an overall interaction only if the complete predeclared program outperforms its matched comparator on durable function, shows the nominated state transitions and causal dependencies, and passes every safety gate. Until then, the architecture is a falsifiable systems hypothesis, not evidence that the entire material set has a favorable cumulative effect.

12. Closing scientific position

Multicomponent repair biology becomes scientifically tractable only after it abandons universal outcome labels and fictitious uniform inputs. The required order is: verify the material; measure the active species and relevant exposure; identify a proximal perturbation; test causal dependency; demonstrate a context-specific function; enforce non-compensatory guardrails; and only then estimate pairwise or higher-order interactions. The present paper provides a curated claim record, bounded mathematical templates, and advancement gates for that sequence. It does not validate a twenty-component intervention. It identifies which narrowly specified preclinical hypotheses deserve to be tested only after their material and evidence gaps are closed.

13. Worked mechanistic reference: mitochondrial membrane state × apoptotic commitment

13.1 Why this is the first demonstration

The first empirical demonstration should be narrow enough to be wrong. A mitochondrial-survival prototype pairs Elamipretide with Humanin under a predeclared mitochondrial stressor. This pairing does not assert a therapeutic combination. It is selected because the agents have distinguishable *candidate* proximal locations: Elamipretide near a cardiolipin-associated inner-membrane state and Humanin near BAX activation. MOTS-c is deliberately excluded from the first factorial model because its metabolic and nuclear-stress behavior adds a third mechanism and a larger identifiability burden.

The preregistered theoretical prediction is sequential complementarity, not assumed synergy: relative to matched single-agent and vehicle conditions, the pair is expected to produce an earlier or more complete correction of the nominated membrane-state measurement, followed by lower BAX/outer-membrane-permeabilization pressure and then a larger improvement in the declared functional-cell outcome. It is equally possible that the pair will be merely additive, redundant, antagonistic, or null. The decisive falsifiers are lack of temporal order, lack of attenuation after membrane or BAX-pathway interruption, no functional interaction under the selected null, or a failed DNA-damage/persistence guardrail.

flowchart LR
    S[Defined mitochondrial stressor] --> C[Cardiolipin / inner-membrane organization]
    C --> A[Energy state]
    A --> B[BAX commitment pressure]
    S --> B
    B --> V[Viable functional cells]
    E[Elamipretide active species] --> C
    H[Humanin active species] --> B
    S --> D[DNA-damage / persistence guardrail]
    D --> V

The diagram is a falsifiable hypothesis. In particular, it can fail if Elamipretide does not alter the selected membrane measurement, if Humanin does not alter BAX pressure, if either functional phenotype persists after the proposed causal node is interrupted, or if increased short-term viability is accompanied by failed guardrails.

13.2 State model for the prototype

Let C(t) be a defined cardiolipin/membrane-organization index, A(t) a named energetic measurement, B(t) a BAX/BAK-associated mitochondrial-commitment index, V(t) the viable functional-cell fraction, and D(t) a DNA-damage/persistence guardrail. Let E(t) and H(t) be measured, normalized active-species exposure functions for Elamipretide and Humanin, and let Z(t) be a calibrated stress input. This is a synthetic design model, not a representation of observed biology:

\frac{dC}{dt}=-k_C(C-C_0)-s_CZ(t)+\theta_EE(t),

\frac{dA}{dt}=-k_A(A-A_0)+\theta_{CA}(C-C_0)-s_AZ(t),

\frac{dB}{dt}=-k_B(B-B_0)+\theta_{AB}\max(0,A_0-A)+s_BZ(t)-\theta_HH(t),

\frac{dD}{dt}=-k_D(D-D_0)+s_DZ(t)+\delta_EE(t)+\delta_HH(t),

\frac{dV}{dt}=-\lambda_{\max}\operatorname{logit}^{-1}(\alpha_0+\alpha_BB+\alpha_DD+\alpha_ZZ(t))V.

All rate constants and theta, s, and delta coefficients have inverse-time units when the indices and exposures are dimensionless. Lambda_max has inverse-time units. The code declares its synthetic parameter units and uses a complete D(t) equation so the guardrail is observable. These relations are not evidence that biological coefficients have the indicated signs in every model. Direct stress effects on viability, BAK-associated alternatives, treatment effects outside nominated nodes, and measurement error remain rival explanations to test.

Y_{nb}\sim\operatorname{Binomial}(N_{nb},p_{nb}),\qquad
\operatorname{logit}(p_{nb})=\alpha+\beta_Eu_{E,nb}+\beta_Hu_{H,nb}
+\beta_{EH}u_{E,nb}u_{H,nb}+a_b,

where a_b is a prespecified block effect. Beta_EH is an interaction on the stated log-odds scale. It is not a universal synergy constant and it cannot be interpreted without a prespecified interaction null, raw-scale contrast, uncertainty interval, and sensitivity analysis.

13.3 Preregistration specification template

The accompanying artifact is a specification template, not a registered protocol. It requires material release, factorial cells, biological unit, block structure, randomization, blinded acquisition, temporal measurements, a declared outcome distribution, primary outcome, guardrails, analysis, and decision rules. It requires a stress-free control for assay interpretation and a stressed four-cell factorial comparison for the interaction estimand. It also requires a pathway-dependency experiment before attribution to the proposed causal route.

The primary functional question is deliberately separate from the molecular measurements: does the combination alter a tissue-relevant viable functional-cell outcome at the stated time? The molecular sequence is supportive only if it exhibits temporal order and pathway dependence. A favorable viability result with persistent DNA damage, senescence/abnormal survival, or another failed guardrail is classified as adverse or uninterpretable.

13.4 Synthetic data demonstration

The companion script parb_mitochondrial_prototype.py generates a completely synthetic blocked four-cell, longitudinal data set. It is a code check and design demonstration, not a model of observed biology. It represents exposure functions, membrane organization, energy charge, mitochondrial-commitment pressure, viability counts, and DNA damage across 0–72 hours. It reports additive-scale and logit-of-condition-mean interaction contrasts with standard errors.

The synthetic example is intentionally useful only as a computational check: the same data-generating process can produce different interaction values on additive and logit scales. It does not establish synergy, effect size, biological plausibility, or power for a real experiment. The self-test verifies bounds, complete factorial/block structure, finite estimates, and seed-stability diagnostics; any real study requires an independently justified simulation and analysis plan.

14. Release artifacts, versioning, and editorial audit trail

This manuscript is accompanied by Revision-3 evidence-recovery artifacts in addition to the retained Revision-2 audit trail:

This draft now releases a claim-level CSV and a multilingual source-recovery record, but it is not a systematic review. Before external publication, the author must supply affiliation, contribution, funding, competing-interest, and material-source statements; select a persistent archive and license; confirm a data/code availability statement; and obtain independent scientific-editor review. A later revision should add a formal scoping-review appendix and independently audited source-extraction table before making any comprehensive-evidence claim.

Appendix A. Bounded dossiers for the tissue-labelled short peptides

This appendix preserves the available observations without converting them into organ-specific pharmacology. It should be read with the identity requirements in Section 3.

A.1 Livagen and Vilon

Livagen and Vilon must remain distinct labels. In a recovered rat organotypic panel, the investigators identify Livagen as KEDA and report a 19±3% increase in liver-explant area index; this is a short, source-specific outgrowth observation, not a direct cell count, hepatic targeting demonstration, or clinical result. [Chalisova et al., organotypic panel](https://fundamental-research.ru/en/article/view?id=36841) Vilon is identified as KE in a human-MSC aging study that reports increased SIRT1 and altered PARP1/PARP2 gene/protein measures in specified aging states. The accompanying DNA-interaction statement rests on molecular modelling rather than direct cellular target engagement. [KE-MSC study](https://pubmed.ncbi.nlm.nih.gov/37782636/)

The sources are not independently replicated outside a limited source network, do not establish a shared receptor or systemic immune pharmacology, and do not verify commercial product-to-sequence mapping. The next experiment for either material is not a combination study: it is a material-controlled concentration-time study in one defined lineage, with direct cell-number, functional, host-defense-guardrail, and early-signal/target-deconvolution endpoints.

A.2 Pancragen and Cartalax

Pancragen is identified as KEDW in the recovered organotypic panel. In rat pancreatic explants, that source reports an 18±3% area-index increase with lower p53 and higher Ki67, FOXA2, and Pax6 measurements. These are an assay-specific outgrowth and marker set, not proof of direct DNA binding, pancreatic uptake, beta-cell specificity, glucose control, or clinical benefit. [Chalisova et al., organotypic panel](https://fundamental-research.ru/en/article/view?id=36841) A small old-rhesus-monkey endocrine record remains hypothesis-generating and source-specific. [Goncharova et al., 2014](https://pubmed.ncbi.nlm.nih.gov/25946840/)

Cartalax is no longer described as identity-free in the literature: NLM indexing maps Cartalax and T-31 to AED. In young and old rat kidney organotypic cultures, AED/T-31 was reported to increase renewal-associated outgrowth and to change Ki67/p53 measures; in a human fetal bone-marrow MSC aging model, AED altered a defined gene-expression panel. [NLM identity record](https://meshb.nlm.nih.gov/record/ui?ui=C572340); [kidney study](https://pubmed.ncbi.nlm.nih.gov/26033601/); [MSC study](https://pubmed.ncbi.nlm.nih.gov/32399807/) This resolves a literature-level nomenclature relationship only. It does not prove commercial product composition, cartilage targeting, matrix restoration, or clinical joint benefit. Neither Pancragen nor Cartalax is eligible for an interaction coefficient before analytical identity, exposure, direct functional, target-dependency, and independent-replication gates.

A.3 Pinealon and Cortagen

Pinealon/EDR has a defined cell-model record: dose-dependent reduction in oxidative-stress-associated ROS accumulation, reduced propidium-iodide-defined necrotic death, delayed ERK1/2 activation, and cell-cycle modification were reported across cerebellar granule cells, neutrophils, and PC12 cells. This is materially stronger than an unqualified “target unresolved” label, but it still does not identify a cognate receptor, prove direct genome interaction, establish brain delivery, or demonstrate human neuroprotection. [Pinealon cellular study](https://pubmed.ncbi.nlm.nih.gov/21978084/) Fluorescence-labelled peptide and oligonucleotide/DNA work supports a biophysical interaction observation, not promoter occupancy or causal transcriptional control in a living target lineage. [Pinealon biophysical study](https://pubmed.ncbi.nlm.nih.gov/22117547/)

Cortagen/AEDP has both a mouse-heart transcript record and a recovered rat organotypic result: AEDP increased cerebral-cortex explant area index by 22±4% versus control in one three-day assay. Neither transcript changes nor this area-index endpoint is a direct target, a pure proliferation measurement, a functional neural result, or evidence of clinical neurobiology. [Cortagen microarray study](https://pubmed.ncbi.nlm.nih.gov/15159690/); [organotypic panel](https://fundamental-research.ru/en/article/view?id=36841)

For both materials, parent-versus-metabolite stability, direct cell-number versus migration analysis, a rapid proximal signal, and a dependency test must precede broad transcriptomics or tissue-specific claims.

A.4 Chonluten and Vesugen

Recovered sources identify Chonluten as H-Glu-Asp-Gly-OH and Vesugen as H-Lys-Glu-Asp-OH in their methods/tables. In a 2024 rat organotypic report, the stated Chonluten sequence increased lung-explant area index by 24±2%; in the same study, the stated Vesugen sequence increased vessel-explant area index by 23±5%. The combinations tested were higher, but the design did not supply a prespecified interaction null, dose surface, or uncertainty sufficient to call synergy. Critically, the paper reverses the names in later discussion prose. Revision 4 uses the methods/tables only and records the conflict rather than treating its narrative mechanism as reliable. [Chalisova et al., 2024](https://intphysiology.ru/index.php/main/article/download/331/236/6723)

For Vesugen/KED, a 2013 conference report used serial chamber counts in human adipose-derived MSC culture and reported +25% versus control at day 3 and a twofold difference at day 5; a separate irradiated-rat study reported PCNA/morphology outcomes in duodenum. The first is a direct short-term cell-count observation, but the report has limited methodological detail and neither study demonstrates generalized MSC expansion, endothelial uptake, vascular targeting, human efficacy, or safety. [Human-MSC conference report](https://szgmu.ru/upload/files/%D0%96%D1%83%D1%80%D0%BD%D0%B0%D0%BB%20%D0%A0%D0%BE%D1%81%D1%81%D0%B8%D0%B9%D1%81%D0%BA%D0%B8%D0%B9%20%D1%81%D0%B5%D0%BC%D0%B5%D0%B9%D0%BD%D1%8B%D0%B9%20%D0%B2%D1%80%D0%B0%D1%87/all-3-2013-low.pdf); [irradiated-rat intestinal study](https://cyberleninka.ru/article/n/reparativnoe-deystvie-peptida-vezugena-na-strukturu-dvenadtsatiperstnoy-kishki-v-modeli-uskorennogo-stareniya.pdf)

These labels remain outside the executable graph until an analytical material record, independent phenotype replication, exposure measurement, and causal target or dependency evidence exist. Their tissue labels may nominate assays; they may not preassign targets or outcomes.

References

Claims and evidence relationships

References

1. A 24-week, double-blind, placebo-controlled study of three dosages of Cerebrolysin in patients with mild to moderate Alzheimer's disease
   Anton Alvarez, Ramon Cacabelos
   European Journal of Neurology, 2006, randomized controlled trial, DOI 10.1111/j.1468-1331.2006.01222.x Open source
2. Thymosin beta4: actin-sequestering protein moonlights to repair injured tissues
   Allan L. Goldstein, Ewald Hannappel, Hynda Kleinman
   Trends in Molecular Medicine, 2005, review, DOI 10.1016/j.molmed.2005.07.004 Open source
3. A rescue factor abolishing neuronal cell death by a wide spectrum of familial Alzheimer's disease genes and Abeta
   Yuichi Hashimoto, Takuro Niikura, Ikuo Nishimoto
   Proceedings of the National Academy of Sciences, 2001, primary research, DOI 10.1073/pnas.101133498 Open source
4. Peptide Bioregulators in Epigenetic Mechanisms of Aging
   Vladimir Khavinson, Boris Kuznik
   Journal of Aging Research, 2014, Source metadata requires independent resolution, source-cited work; identity unresolved
5. The mitochondrial-derived peptide MOTS-c promotes metabolic homeostasis and reduces obesity and insulin resistance
   Changhan Lee, Jennifer Zeng, Brian G. Drew, Pinchas Cohen
   Cell Metabolism, 2015, primary research, DOI 10.1016/j.cmet.2015.02.009 Open source
6. Marshaling the Anti-Inflammatory Influence of the Neuroimmunomodulator alpha-MSH
   J. M. Lipton, A. Catania
   News in Physiological Sciences, 2000, source citation differs from indexed record, DOI 10.1152/physiologyonline.2000.15.4.192 Open source
7. Regenerative and Protective Actions of the GHK-Cu Peptide in the Light of the New Gene Data
   Loren Pickart, Anna Margolina
   International Journal of Molecular Sciences, 2018, review, DOI 10.3390/ijms19071987 Open source
8. ARA 290, a peptide derived from the tertiary structure of erythropoietin, produces long-term relief of neuropathic pain coupled with suppression of the spinal microglia response
   Maarten Swartjes, Michael Brines, Anthony Cerami, Albert Dahan
   Molecular Pain, 2014, primary research; source citation differs from indexed record, DOI 10.1186/1744-8069-10-13 Open source
9. Targeting mitochondrial cardiolipin and the cytochrome c/cardiolipin complex to promote electron transport and optimize mitochondrial ATP synthesis
   Hazel H. Szeto
   British Journal of Pharmacology, 2014, primary research; source citation differs from indexed record Open source
10. Pentadecapeptide BPC 157 and the Angiogenesis in Healing and Tumor Models
    Jaksa Vukojevic, Marija Milavic, Darko Perovic, Predrag Sikiric
    Current Issues in Molecular Biology, 2022, source-cited work; identity unresolved

Available representations

• html (text/html; charset=utf-8)
• json (application/json; charset=utf-8)
• receipt (application/json; charset=utf-8)
• artifact (application/pdf), PARB_Revision_4_PreRelease.pdf