Correlation Architecture · Cross-Domain Methodology · Independent Research
Inner Architecture LLC develops and validates a correlation-architecture methodology — weighted Jaccard on continuous correlation vectors — for detecting network reorganization in biological, industrial, financial, ecological, and neuroscientific systems. Components first. Relationships between components as the measured signal. Categories as findings, not inputs.
About
An independent research organization developing a domain-invariant methodology for measuring how correlation architectures reorganize when complex systems change state.
Inner Architecture LLC is an independent research organization headquartered in Canton, Ohio. The company's core scientific contribution is the development and cross-domain validation of weighted Jaccard (WJ) similarity on continuous correlation vectors as a general tool for quantifying network reorganization — how much of a system's pairwise relationship structure is preserved between conditions, and how much is restructured.
The methodology is designed around a single governing principle: relationships between components carry information that components alone do not. For an N-component system, pairwise relationships grow as N(N−1)/2 and the full multi-way configuration space grows faster still. Conventional monitoring collapses this relational information into component-level aggregates and loses the architectural signal. WJ preserves it, and detects reorganization earlier than threshold-based or variance-based alternatives across every domain we have tested.
The research program currently comprises 20 manuscripts across 15 target journals, spanning computational genomics, clinical transcriptomics, industrial predictive maintenance, financial regime detection, ecological tipping-point detection, functional brain connectivity, seismology, structural health monitoring, cybersecurity, quantum hardware characterization, and climate dynamics. Six papers have been published through peer review in 2026: a genome-wide co-expression analysis showing that epigenetic regulators are preferentially coordinated with the clustered protocadherin gene family across the human brain (Frontiers in Genetics); an analysis of divergent Sigma-1 and Sigma-2 receptor co-expression architectures in human brain (Frontiers in Pharmacology); a seven-cohort meta-analysis of ALDH1A1 co-expression network reorganization across Parkinson's disease substantia nigra cohorts (Frontiers in Aging Neuroscience); a threshold-free multivariate indicator of correlation-network reorganization validated across coral reef, freshwater lake, and climate teleconnection systems (Ecological Indicators); a pairing-family decomposition of EIF2S1 and PELO co-expression architecture in human brain (Frontiers in Molecular Neuroscience); and an analysis showing that functional connectivity reorganization after propofol anesthesia exceeds within-state measurement variability and is only marginally reversed during early recovery (NeuroImage: Reports). Additional manuscripts are currently under peer review or in revision, including a Hypothesis and Theory article on sigma-1 dual-modality activation in fifth-round revision at Frontiers in Pharmacology and a multi-cohort PTSD blood-transcriptomic analysis under revision at the Journal of Psychiatric Research.
Inner Architecture holds European Genome-Phenome Archive institutional access, is registered as a federal contractor on SAM.gov, and operates a full methodology-governance architecture — auditors, validators, and execution-trace memory — that enforces WJ methodology compliance at the analytical level before any pipeline runs.
Drake Harbert is the architect of the WJ methodology and sole author on six peer-reviewed publications and on every manuscript in the Inner Architecture pipeline. His work integrates genome-wide co-expression analysis, multivariate industrial sensor monitoring, functional neuroimaging, and cross-domain methodological validation. He serves as a peer reviewer for the Journal of Psychopharmacology (SAGE), and holds a provisional patent on endogenous DMT biosensor technology. ORCID: 0009-0007-7740-3616.
Peer-Reviewed Acceptances
Protocadherin / epigenetic-regulator coordination (Frontiers in Genetics) · Sigma-1 / Sigma-2 receptor divergence (Frontiers in Pharmacology) · ALDH1A1-dopaminergic Parkinson's meta-analysis (Frontiers in Aging Neuroscience) · WJ correlation-network reorganization across coral, freshwater and climate systems (Ecological Indicators) · EIF2S1-PELO translational quality-control architecture (Frontiers in Molecular Neuroscience) · Connectivity architecture after propofol anesthesia (NeuroImage: Reports)
Peer Review Service
Invited reviewer for SAGE-published pharmacology research; completed review of a serotonin-receptor response study
Institutional Access
Controlled human genomic dataset access granted to Inner Architecture LLC as a registered institution
Federal Registration
Federal contractor registration enabling NIH, NINDS, and DOD grant applications
Primary Data Sources
Cross-domain validation spans public genomics repositories, industrial benchmark datasets, and openly-released human neuroimaging archives
Governance Architecture
Pipeline validators, compliance auditors, provenance tracking, and execution-trace memory enforce methodology before any analysis runs — a structural safeguard against domain-conventional drift
Intellectual Property
Provisional patent filed for endogenous DMT biosensor technology; documented WJ methodology framework deployed across nine validated research domains
Research & Publications
Six papers published through peer review in 2026, within a pipeline of twenty manuscripts across fifteen target journals, all unified by the weighted Jaccard correlation-architecture methodology. Featured work below; the full pipeline follows.
NeuroImage: Reports · accepted September 2026
Whole-brain correlation architecture across wakefulness, propofol-induced unresponsiveness and early recovery in 24 subjects. Establishes the within-state reference that similarity indices require, showing that repeated measurement of a single state returns 0.544 rather than 1.0, and correcting the state-attributable share of reorganization to 11 to 14 percent. Reorganization is redistributive rather than uniform, so a scalar summary of total connectivity cannot distinguish wakefulness from recovery where the full-matrix comparison can. Group averaging is shown to inflate apparent recovery, with subject-level recovery of 3.2 percent against a group-level estimate an order of magnitude larger.
Frontiers in Molecular Neuroscience · doi:10.3389/fnmol.2026.1817096
EIF2S1 and PELO share near-identical genome-wide co-expression architecture (continuous weighted Jaccard = 0.914) while diverging at their functional tails, separating a constitutive transcriptional coupling layer from a functional interaction layer in the ribosome-associated quality control system.
Frontiers in Aging Neuroscience · doi:10.3389/fnagi.2026.1806505 · publishing May 2026
Seven-cohort random-effects meta-analysis (n = 156; 70 control, 86 PD) demonstrating disease-selective attenuation of ALDH1A1's co-expression with dopaminergic pathway genes in substantia nigra. ALDH1A1-dopamine correlations fall from a control mean r = 0.92 to r = 0.54 in PD (mean Δr = −0.336), while intra-dopaminergic correlations attenuate less than half as much. Permutation testing establishes selectivity at p = 0.0052. Category-level attenuation partitioning identifies ALDH1A1-associated coordination as the primary site of disease signal.
Neurobiology of Disease · GTEx v8 · 5 brain regions · 14,284 genes
WJ applied to genome-wide co-expression vectors of three ribosome-associated quality control (RQC) factors across five brain regions. Hippocampus shows elevated cross-region reorganization for LTN1 (+0.060, p = 0.030) and NEMF (+0.050, p = 0.059), but not PELO. Partial correlation controlling for six cell-type composition proxies preserves the LTN1 hippocampus effect (+0.033, signed p = 0.005) but eliminates the NEMF effect, dissociating cell-intrinsic from cell-composition-dependent regional regulation. Five negative-control genes show no comparable effect. Signed-versus-unsigned WJ divergence of 0.068 indicates roughly 30% of regional reorganization arises from correlation-sign inversions that binary overlap cannot detect.
Frontiers in Pharmacology · Neuropharmacology section · GTEx v8 · n = 209 · 16,225 genes
Three WJ formulations — shifted, unsigned, and signed — applied to SIGMAR1 and TMEM97 genome-wide correlation vectors reveal that the two receptors share the majority of their global transcriptional architecture (WJ shifted = 0.964, unsigned = 0.907, signed = 0.906; rank-identical across 21 pairwise comparisons, ρ = 1.000). Their top-5% co-expression networks, however, overlap by only 10.0% (binary Jaccard = 0.100). The 2.1-fold variation in WJ-versus-binary dissociation gap tracks known biological relatedness rather than a fixed methodological property. Distinct GO enrichments — SIGMAR1 mitochondrial translation and TCA cycle; TMEM97 ubiquitin-mediated proteolysis and neurodegeneration — provide transcriptomic rationale for subtype-selective pharmacology.
Frontiers in Genetics · Article 1807347 · GTEx v8 · 2,642 brain samples · 13 brain regions
Unbiased genome-wide co-expression analysis ranking every gene by its co-expression with the clustered protocadherin family (PCDHA, PCDHB, PCDHG) across 13 human brain regions. Epigenetic regulators — chromatin remodelers, DNA-methylation machinery, and histone modifiers — are preferentially co-expressed with protocadherins region-wide, extending the well-characterized locus-specific epigenetic control of stochastic protocadherin isoform choice to a broader shared transcriptional program. The effect replicates in an independent cohort (GSE80655) and survives cell-type deconvolution, indicating transcriptional rather than compositional coordination. A substantial fraction of the coordinated regulators are themselves genes whose loss of function causes neurodevelopmental disorders (ARID1A, EP300, CREBBP, CHD8, MECP2, KMT2A, KMT2D), connecting protocadherin transcriptional context to chromatin-disorder biology.
Ecological Indicators · doi:10.1016/j.ecolind.2026.115024
Application of weighted Jaccard similarity on continuous pairwise correlation vectors to long-term water-quality monitoring data from lake systems undergoing regime shifts between clear-water and turbid-water states. The architectural relationships between water-quality parameters reorganize before any individual parameter crosses its threshold, producing an early-warning signal that conventional variance- and autocorrelation-based methods (in the Scheffer lineage) do not surface as cleanly. Demonstrates that the cross-domain WJ methodology validated in genomic and industrial substrates also detects pre-transition reorganization in noisy ecological systems — generalizing the architectural-failure-precedes-component-failure principle to a fourth physical substrate.
NeuroImage · OpenNeuro ds006623 · n = 24 · 456 ROIs
Subject-level WJ applied to full pairwise Spearman correlation matrices across 456 cortical regions during wakefulness, propofol-induced unconsciousness, and early recovery. All 24 subjects show significant architectural similarity reduction under anesthesia (mean WJ = 0.499, SD = 0.043; Wilcoxon p < 0.001), with per-subject connectivity preserving approximately half of its weighted edge structure. Recovery is minimal: mean recovery-phase WJ = 0.522, per-subject recovery fraction = 3.8% (p = 0.021). All eight canonical networks show significant reduction (FDR-corrected); only the default mode network shows significant recovery. Propofol concentration at loss of responsiveness correlates with dissimilarity magnitude (ρ = −0.591, p = 0.002).
6 published · 20 manuscripts · 15 target journals · 9 validated domains
Research Focus Areas
Biomedical Genomics
Correlation-network reorganization as a primary disease signal in Parkinson's, PTSD, ALS-associated pathways, and sigma-receptor biology
Functional Neuroscience
Subject-level connectivity-architecture similarity during anesthesia, unconsciousness, and cross-scale gene-to-brain correspondence
Clinical Translation
Blood-transcriptomic architecture reorganization in sepsis, trauma, and stress-related network decoupling
Industrial Systems
Peer-reviewed research on sensor-correlation reorganization as an interpretable unsupervised health indicator — validated on NASA C-MAPSS, Scania, and Tennessee Eastman benchmark datasets
Physical Sciences
Pre-transition correlation reorganization in climate regimes; seismic-station networks; qubit-calibration drift detection
Quantitative Systems
Regime-change detection in individual-stock networks, ecological parameter tipping points, and network-flow correlation reorganization
Methodology
A domain-invariant methodology for quantifying how much of a system's correlation architecture is preserved between conditions, and how much reorganizes.
Given any two conditions A and B of a system with N components (genes, sensors, neural regions, stocks, stations), compute the full N(N−1)/2 pairwise correlation matrix under each condition. WJ compares the two correlation vectors as a continuous similarity measure — no binarization threshold, no pre-selection of “significant” edges, no dependence on component-category labels. The result is a single scalar describing how much of the relational architecture survives the transition.
WJ(a, b) = Σ min(|ai|, |bi|) / Σ max(|ai|, |bi|)
Every pipeline computes unsigned and signed variants in parallel. Their divergence directly measures the fraction of reorganization arising from correlation-sign inversion — structural signal that binary edge-overlap metrics cannot detect.
Identify the smallest individual measurement in the domain — individual gene, individual sensor, individual region. Never a category, subsystem label, or pre-aggregated bundle.
Compute the complete pairwise correlation matrix across all fundamental units. No pre-filtering. No hypothesis about which pairs should relate. Categories emerge from the architecture, not from assumption.
Apply unsigned and signed WJ to compare correlation vectors across conditions, timepoints, or regimes. Report implementation divergence — the sign-inversion fraction — as a first-class output.
Permutation nulls with 1,000+ iterations, FDR correction across all pairs, and structure-aware null models (Mantel permutation, spin tests) where matrix-derived pairs create dependence.
Intellectual Property
Provisional patents and methodology disclosures arising from the research program.
Provisional Patent Filed · 2025
Real-time biosensor system for measuring endogenous N,N-dimethyltryptamine concentrations in biological fluids. Leverages sigma-1 receptor binding affinity characteristics for physiological DMT detection with applications in diagnostic monitoring, research instrumentation, and therapeutic guidance.
Non-invasive measurement of endogenous DMT levels as a biomarker for neurovascular function
Laboratory detection platform for quantifying DMT in tissue samples, blood, and cerebrospinal fluid
Real-time feedback during pharmacological interventions targeting sigma-1 receptor pathways
Miniaturizable sensor architecture for clinical deployment in neurology and psychiatry
Contact
Inner Architecture LLC welcomes correspondence from academic researchers, journal editors, and peer reviewers on the published and under-review work documented above. The research program operates as an independent publication vehicle; methodology and reference implementations are documented in peer-reviewed venues and on the linked GitHub repository.
Currently reviewing for the Journal of Psychopharmacology (SAGE); available for methodological review on cross-domain correlation-network manuscripts
Research groups extending WJ methodology to new domains; data-sharing arrangements for methodology validation on additional substrate classes
Editors and journal correspondents regarding the manuscripts listed above or upcoming submissions across the framework family