Research & collaboration
Adaptive BioAge is operational and has a public architectural specification; its exact frozen reproduction package is in preparation, and the model has not yet been independently externally validated. BioMIR seeks partners for retrospective external validation, prospective repeatability and responsiveness, and preventive-health workflow studies.
Priority validation questions
These are study questions, not established product claims. Designs that can falsify, calibrate, or constrain the model are especially useful.
Does within-person change in the higher-frequency Adaptive BioAge trajectory show meaningful directional convergence with episodic KDM and Levine PhenoAge, and can it identify sustained physiological drift between clinical assessments?
Does presenting complex longitudinal physiology as an intuitive Δ-year trajectory with contributor-level attribution improve awareness of modifiable risk factors, sustained engagement, preventive behavior, and clinically relevant risk-factor trajectories?
How stable are daily and domain-level estimates under repeated comparable input observations, missingness, device variability, carried observations, and plausible changes in model parameters?
Can users and clinicians distinguish short-term fluctuation from persistent trajectory, identify the contributors that explain a change, and determine whether the framework adds information beyond conventional risk factors or isolated readings?
Can downstream on-device or federated models improve individualized calibration, temporal-response characterization, or action prioritization beyond the locked deterministic baseline while preserving reproducibility, privacy, and a clear boundary between learned interpretation and authoritative scientific computation?
Can a bounded semantic and language-model layer explain deterministic BioMIR results with numerical fidelity, explicit provenance, correct handling of missing or stale data, and clear separation of observation from interpretation—while remaining optional to normal BioMIR computation?
Collaboration fit
Repeated wearable, home-monitoring, or Apple Health-compatible data paired with clinical biomarkers or outcomes.
Calibration, sensitivity analysis, missing-data strategy, temporal mediation, within-person modeling, subgroup performance, and uncertainty communication.
Longitudinal modeling, on-device learning, individualized calibration, federated learning, secure aggregation, privacy-preserving machine learning, and digital-biomarker methods that can test whether downstream personalization adds reproducible value beyond the locked deterministic BioMIR baseline without centralizing raw longitudinal health records.
Canonical scientific knowledge representation, typed semantic APIs, tool-calling fidelity, provenance-linked explanation, on-device language-model evaluation, voice and narration, App Intents/Siri-facing interaction, hallucination testing, privacy/data minimization, and human-factors methods for constrained conversational health intelligence.
Cardiometabolic prevention, aging research, longitudinal physiology, digital biomarkers, and pragmatic interpretation of multi-timescale health data.
Prospectively specified evaluation of BioMIR outputs as exploratory longitudinal analytics or protocol-defined endpoints where appropriate. Adaptive BioAge should not be represented as a validated surrogate or evidence of treatment benefit without fit-for-purpose validation.
Testing whether contributor-level explanations and data-quality signals improve understanding without creating false precision or inappropriate clinical inference.
Lock or explicitly version the algorithm, reference data, parameter sets, and software for a study, and prespecify provenance, access, consent or ethics review, analysis responsibilities, publication policy, and permitted reuse before nonpublic participant-level data are exchanged.
Participant-controlled data
BioMIR is adding a Settings export that lets users select a domain and generate a de-identified longitudinal dataset of relevant BioMIR measures and model outputs. The export is created under user control for independent N-of-1 analysis or voluntary contribution to an eligible study. Generating an export does not transmit data to BioMIR, enroll the user in research, or authorize secondary use.
Under a governed research protocol, independently contributed longitudinal records can be analyzed as N-of-1 series or cohorts to test parameter stability, between-person heterogeneity, transportability, and responsiveness. These data may inform candidate parameter refinements; they do not automatically retrain or alter the production model.
Any proposed parameter revision should be prespecified, versioned, evaluated on held-out or time-forward data where feasible, and independently validated before promotion into a production model. Formal participation requires its own protocol, consent process, data-governance plan, and applicable ethics review or determination.
Study responsibilities
The objective is independent evidence, not a promotional partnership. Responsibilities should be defined before analysis so model development, data custody, statistical decisions, and interpretation remain traceable.
BioMIR would provide
Research partner would provide
Agreed jointly before analysis
BioMIR pairs a modeled age-equivalent contribution with a longitudinal persistence index: the proportion of qualified modeled days in High or Highest tiers, with fitted trends across prespecified periods. This is descriptive context, not disease probability or a validated clinical threshold. Independent studies must test repeatability, responsiveness and incremental information beyond the latest measurement and conventional averages, while separating held inputs from original observations and accounting for correlated repeated days. A 9 October 2026 draft amendment to the working v1.3 protocol specifies analytical checks, independent participant/time-forward evaluation, freshness sensitivity and human-factors endpoints; clinical and biostatistical review, preregistration and study execution remain pending.
The study plan freezes tiers, windows, denominator and version lineage. Clinical utility and any future referral thresholds require separate evidence and approval. Research collaborators can propose independently measured endpoints, acquisition schedules and participant-level external evaluation designs.