Source Code and Dataset Supporting Reproducibility of Results for "Finite Prevention Windows for HIV Post-Exposure Prophylaxis: Irreversible Proviral Integration Defines Route-Specific Intervention Limits" (Science Advances submission ID aeh5879).

This is the v2 reproducibility archive for the manuscript "Finite Prevention Windows for HIV Post-Exposure Prophylaxis: Irreversible Proviral Integration Defines Route-Specific Intervention Limits" (Science Advances submission ID aeh5879). The v2 deposit responds to a self-disclosed editorial correction (disclosure tag 2026-04-30, commit d047d2d) clarifying that the v1 manuscript's reported t_crit values corresponded to the model's hand-coded seeding_midpoint parameters rather than to t_crit(η) as defined by Theorem S3.1(iv). The v2 implementation replaces the prior phenomenological logistic model with a multiscale within-host implementation: - Phase 1: stochastic tau-leaping at low founder population (V_0 ≤ 10), capturing extinction events - Phase 2: deterministic ODE handoff once productive infection is established - Eclipse phase: discrete fixed-delay buffer (~22 h, Perelson 1996 Table 2) Heterogeneity is implemented as multiplicative log-normal noise (CV = 0.3) on {β, c, δ, α}. Verified numerical results at the canonical operating point (V_0 = 10^3 parenteral, V_0 = 1 mucosal, CV = 0.3, η = 0.05): - Parenteral t_crit ≈ 34.5 hours - Mucosal t_crit ≈ 60.5 hours - Compression ratio ≈ 1.75× (concordant with NHP empirical 1.5–2× range) Reproducibility: all numerical claims in the manuscript reproduce from this archive at the cited git SHA. The archive includes: - SRC/multiscale_model/: multiscale within-host model implementation (multiscale_v3.py, run_mc_v3.py) - SRC/multiscale_model/results/, SRC/multiscale_model/results_v3/: simulation outputs (CSVs) - core_theorem/, route_models/, stochastic_layers/: supporting modules - aidsvu datasets/, data/: input surveillance data - v2_revision/: full Science Advances v2 submission package (LaTeX manuscript, supplement, cover letter, figures, README, numerical_claims_v2.csv) - audit_check.py: original audit diagnostic that identified the v1 t_crit conflation - archive/root_archive/audit_verification_output_2026-04-26.txt: committed audit output Run instructions are in README.md and v2_revision/README_v2.md. Reproduce verified t_crit values with: cd SRC/multiscale_model && python3 run_mc_v3.py Pinned package versions: numpy 1.26.4, scipy 1.12.0, matplotlib 3.9.4, pandas 2.3.3 (see requirements.txt). This deposit preserves the disclosure-2026-04-30 tag (at commit d047d2d) unchanged; the v2-submission tag annotates the corrected package state at submission.