A generator-matrix causal-inference framework separates measurable aging biomarkers from mortality-driving latent dynamics in humans — code and derived data

Analysis/figure-generation code and derived (aggregate) result tables for the study "A generator-matrix causal-inference framework separates measurable aging biomarkers from mortality-driving latent dynamics in humans" (Tanigawa & Iwaki, 2026; manuscript submitted). The study asks, in one framework, how much of human mortality acceleration is captured by measurable molecular aging markers, whether the measurable part is causal, and whether it is reversible, using three orthogonal methods: (1) a Bayesian Markov generator-matrix model of hallmark-load dynamics with death as an absorbing state, fitted to NHANES (with linked mortality) and replicated in HRS; (2) two-platform cis-pQTL Mendelian randomization plus colocalization (UKB-PPP, deCODE) against parental-lifespan GWAS, calibrated with known-causal positive controls; and (3) a chronological-vs-causality-enriched damage clock battery on cellular-reprogramming datasets. Contents: all analysis and figure-generation scripts (Python; some R/shell), derived/aggregate result tables (CSV) with per-analysis findings notes, pinned dependency files (environment/), DATA_PROVENANCE.md, LICENSE.md, and CITATION.cff. The manuscript text and rendered figures are NOT included (available via the journal article). Data governance: this record does NOT redistribute restricted or licensed primary data (HRS individual-level data; UKB-PPP and deCODE pQTL summary statistics; GEO methylation matrices). HRS-derived tables are aggregate only (cell sizes >= 5). Obtain primary data from the providers as described in DATA_PROVENANCE.md. Licensing: this record is released under CC BY 4.0; code in code/ is additionally licensed under the MIT License (see LICENSE.md). Funding: JSPS KAKENHI Grant Number JP24K15812.