Day 205 on ruxolitinib · day 182 on a beta-blocker · 1048 days post-transplant · Oura ring · data through Oct 06 2026

I had a stem-cell transplant for leukaemia in November 2023. This site reads my Oura ring every night and turns the data into the reports below. Claude Code wrote the code; I described what I needed in Norwegian and checked what came back. Open source, MIT licensed, part of TEEI Health Equity. Companion project: Genomics portal. How this was built.

The digital twin

A model projected my body without the medicines. I went far past it.

This twin learned my nightly physiology from the 67 nights before treatment and projects the range my HRV would sit in without it: around 10 ms. On ruxolitinib alone, 1 of 23 nights sat above that range. After a beta-blocker was added on April 8, 173 of 182 did. The gap between the two lines is what a consumer ring made visible; the phase analysis below says which medicine carries it.

Actual, on treatment Modelled without treatment Projected range without treatment

173 of 182 nights since the beta-blocker sit above the range the twin projected without treatment, against 1 of 23 on ruxolitinib alone. Projected 10 ms. Actual 24 ms, a +139% gap.

Phase-corrected estimate (piecewise ITS with AR(1) errors): ruxolitinib step -1.5 ms (p=0.71), beta-blocker step +19.1 ms (p<0.001). Tau-U, trend-corrected: pre to ruxolitinib alone -0.19; ruxolitinib alone to plus beta-blocker +0.95.

Whole-period test: HRV +164% (p<0.001), all pre-ruxolitinib nights vs all post-ruxolitinib nights. In this pipeline's own placebo calibration that test fires at 100% of dates where nothing happened, so it is shown as description, not evidence. Placebo tests · Piecewise ITS · Tau-U.

What moved, and by how much

Last 30 nights vs the pre-treatment baseline. Every figure is a live query.

HRV (RMSSD)

9.0 → 23.1 ms

+155%

Parasympathetic, vagal, autonomic tone. Higher is better recovery.

Average sleeping HR

85 → 67 bpm

-21%

Lower resting heart rate means less cardiovascular strain.

Lowest nightly HR

77 → 59 bpm

-23%

The floor heart rate reaches overnight.

REM sleep

0.88 → 1.33 h

+51%

The sleep stage tied to memory and emotional recovery.

All reports

Core4
Full AnalysisHeart rate, HRV, sleep, activity, SpO2, and readiness trends across the full observation window.Biomarker TrendsComposite biomarker indices combining multiple Oura signals into research-use summary scores.Sleep AnalysisSleep architecture, staging distribution, efficiency, and circadian rhythm analysis.Weekly TrackerOne-page weekly tracker with watchpoints, week-over-week deltas, and clinician-style summary text.
Clinical6
Causal: RuxolitinibBayesian causal impact and interrupted time-series analysis of ruxolitinib response.GvHD PredictionHidden Markov and state-space models predicting GvHD flare probability from wearable signals.SpO2 & BOSSpO2 trend monitoring and bronchiolitis obliterans syndrome screening thresholds.Rux ForecastNear-term HRV and heart-rate recovery forecast from the current post-treatment trajectory.Research SynthesisTwo-hit autonomic recovery hypothesis with live KPIs, timeline, and statistical evidence.Treatment ReportPrimary clinical report covering all systems and both medicines for specialist review.
Advanced5
Advanced HRVFrequency-domain HRV, Poincare plots, DFA, sample entropy, and autonomic balance metrics.Digital TwinUnscented Kalman Filter digital twin tracking latent inflammatory and autonomic states.Foundation ModelChronos foundation model forecasting with prediction intervals and anomaly scoring.Anomaly DetectionMatrix Profile, Isolation Forest, and CUSUM anomaly detection across biometric channels.3D DashboardInteractive 3D scatter of sleep, HRV, and activity with treatment phase coloring.
Comparative8
Autonomic ComparisonHRV and resting HR recovery trajectories compared between post-HSCT and post-stroke patients.Treatment ResponseChangepoint detection and pre/post treatment response with Mann-Whitney U tests.Sleep ArchitectureSleep architecture, efficiency, and timing compared against clinical benchmarks.Activity-RecoveryActivity-recovery coupling analysis: does day N activity predict day N+1 recovery?Anomaly PatternsAnomaly fingerprinting and clustering: how bad days manifest differently.Breathing AnalysisRespiratory-rate trends, week-over-week shifts, and outlier nights against recent baseline.Temperature AnalysisTemperature deviation tracking, excursion alerts, and post-treatment change patterns.P2 ChangepointsPatient 2 changepoint scan of HRV, sleep, and recovery markers around key timeline events.
Statistical4
Piecewise ITSPiecewise ITS regression with AR(1) errors: two-intervention model with date sensitivity analysis.Sequential CISequential Bayesian CausalImpact isolating Jakavi and beta-blocker effects in separate runs.Placebo TestsFalsification tests at 20 random pre-treatment dates to calibrate false positive rates.Tau-U EffectsTau-U and NAP effect sizes for single-case experimental design with baseline trend correction.
Individual2
P2 DashboardPost-stroke patient P2: HRV, HR, sleep, and activity dashboard.P3 DashboardHealthy control P3: baseline HRV, HR, sleep, and activity reference.
Context4
AboutMethodology, limitations, and honest assessment of what this system can and cannot do.Next StepsPlanned analyses, validation targets, and next steps for the digital twin platform.How this was builtWho did what, how a number gets onto a page, and how the pipeline checks itself. Measured, not claimed.Every number, checkedEvery statistic printed on this site, cross-checked against the JSON the pipeline computed it from.