Noise Correlation Length Distinguishes Neurometabolic Protection from Vulnerability Across HIV Infection Phases

Bayesian analysis code, consolidated neuroimaging dataset, and full reproducibility suite for "Noise Correlation Length Distinguishes Neurometabolic Protection from Vulnerability Across HIV Infection Phases" (Demidont 2026, bioRxiv: https://doi.org/10.64898/2026.02.10.703895).

This repository implements a hierarchical Bayesian framework using PyMC to test whether environmental noise correlation length (ξ) distinguishes preserved from vulnerable neurometabolic states across HIV infection phases. The framework is validated across five complementary computational models including hierarchical Bayesian inference, enzyme kinetics modeling, and WAIC/LOO model comparison. All manuscript figures and statistical results are reproducible from a single command (make reproduce).

─────────────────────────────────────────────────────────────PRIMARY DATASET (new in v3.0)─────────────────────────────────────────────────────────────MASTER_HIV_MRS_DATABASE_v3_CORRECTED.csv is the first publicly available open-source consolidated dataset of phase-specific HIV neuro-metabolic MRS data. It contains 49 observations across 7 independent human cohorts (Sailasuta 2012, Young 2014, Sailasuta 2016, Mohamed 2010, Valcour 2012/2013, Chang 2002, Dahmani 2021 meta-analysis) plus NHP reference data (Ratai 2009/2010, SIV rhesus macaque), spanning four decades of published literature (1985–2021) across two continents and two ART eras.

Each row is fully annotated with:- Phase_Canonical: harmonized phase label (acute/chronic/control)- Measurement_Type_Reported: whether NAA/Cr was directly reported or   derived from absolute mM via standard ratio conversion (NAA_mM ÷ Cr_mM)- Ratio_Derivation: complete derivation methodology per study- Included_In_Model: Yes / No / Holdout, with documented exclusion reasons- Source_DOI: primary publication DOI for every row

Complete column definitions, phase definitions, inclusion/exclusion criteria, and derivation methodology are in DATA_DICTIONARY.md.

No open-source consolidated database of phase-specific HIV neuro-metabolic MRS data previously existed. This dataset enables direct replication and extension of the present analysis and constitutes an independent scientific contribution to the HAND neuroimaging literature.

─────────────────────────────────────────────────────────────KEY FINDINGS─────────────────────────────────────────────────────────────ξ_acute = 0.425 ± 0.065 nm vs ξ_chronic = 0.790 ± 0.065 nmP(ξ_acute < ξ_chronic) = 0.914 (non-overlapping 95% HDI)Protection exponent β_ξ = 2.33 ± 0.51 (superlinear scaling)Model-independent signal: phenomenological and enzyme kinetics models converge on identical parameter estimates

──────────────────────────────────────VERSION HISTORY─────────────────────────────────────────────────────────────v3.0 (2026-03-31): Added MASTER_HIV_MRS_DATABASE_v3_CORRECTED.csv and DATA_DICTIONARY.md as primary dataset files. Updated manuscript to bioRxiv v3 (post-hoc corroboration from Zhan 2024, Ma 2025, Zhou 2024, Qi 2023; white matter FA predictions; developmental gradient predictions; cross-species evolutionary framework). Added companion paper reference (10.1101/2025.06.08.658222).

v2.0 (2026-02-18): Corrected supplementary materials (five caption/table fixes); reorganized archive; removed IDE artifacts.

v1.0 (2026-02-13): Initial deposit.

Preprint: https://doi.org/10.64898/2026.02.10.703895 (bioRxiv)

GitHub: https://github.com/Nyx-Dynamics/noise_decorrelation_hiv

Interactive summary (narrated slide deck, mind map, infographic): https://nyxdynamics.org/research/noise-decorrelation/