ELPAC: Endpoint-Anchored Latent Progression with Stage-Varying Multimodal Coordination
Organizations: Department of Biostatistics, Yale School of Public Health, Yale University New Haven, CT 06510
Abstract
Aging cohorts increasingly combine imaging, plasma, and cognitive measurements, but most subjects provide only a partially observed cross-sectional snapshot, leaving latent stage, progression subtype, and stage-varying multimodal coordination to be inferred jointly. Existing methods estimate the cohort-level aging process, model covariance under observed covariates, or fuse modalities for prediction. However, they rarely target the joint problem of endpoint-anchored staging and progression-resolved residual coordination across modalities, even though subject staging, subtype assignment, and multimodal interplay are mutually dependent. We introduce ELPAC, a Bayesian framework with two interlocking constructions. The first integrates subtype trajectories backward from clinically reliable endpoint groups along biologically structured velocity components encoding modality-specific monotone direction and optional sequential ordering priors. The second factorizes residual multimodal variation around the inferred stage axis into sparse coordination axes whose loadings and activations expose which multimodal patterns are operative in each subtype and when they emerge, with stage-resolved variation supplying the auxiliary structure for conditional identifiability. We use a scalable amortized variational inference scheme that enables joint posterior inference at cohort scale with single-pass deployment on new subjects. On synthetic data and a harmonized ADNI+NACC/SCAN Alzheimer's cohort, ELPAC demonstrates accurate latent-structure recovery, well-calibrated prediction intervals, and biologically interpretable multimodal coordination programs that differ by subtype.
Figures & tables
| Symbol | Meaning | Symbol | Meaning |
|---|---|---|---|
| ; | subtype; progression stage | ; ; | no. subtypes, biomarkers; modality of |
| ; | biomarker vector; observed entries | ; ; | covariates; cognition; its stage trend |
| ; ; | anchors; healthy reference; unlabeled | ; | anchor; healthy-reference stage location |
| ; | mean trajectory; within-stage covariance | ; | subtype endpoint; shared healthy origin |
| ; ; | shell state; shell velocity; no. shells | ; | modality direction; modality onset center |
| ; ; | inclusion; onset timing; onset sharpness | ; , | signed velocity magnitude; spline basis, size |
| Method | NLL/cell | BAcc | Macro-F1 | Stage-score | AUC {1,2}/3 | Masked NLL | Cog |
|---|---|---|---|---|---|---|---|
| SuStaIn | – | – | |||||
| DEBM | – | – | – | – | |||
| Static-FA | – | – | |||||
| MoPoE-VAE | |||||||
| XGBoost | – | – | |||||
| Label-prop | – | – |
| Model | Held-out NLL | BAcc | Stage-score | Masked NLL | Stage-binned PPC | Cog |
|---|---|---|---|---|---|---|
| Elpac , no-anchor | ||||||
| Diagonal covariance | – | |||||
| Static-factor covariance | ||||||
| Elpac (full) |
| Method | NLL/cell | RMSE | MMSE | Macro-F1 | BAcc | AUC AD/(CN+MCI) | AUC AD/MCI |
|---|---|---|---|---|---|---|---|
| SuStaIn | – | – | |||||
| DEBM | – | – | – | – | |||
| Static-FA | – | – | |||||
| MoPoE-VAE | |||||||
| XGBoost | – | – | |||||
| Label-prop | – | – |
| Model | Held-out NLL | BAcc | AUC AD/(CN+MCI) | AUC AD/MCI | Masked NLL | Cog |
|---|---|---|---|---|---|---|
| Elpac , no-anchor | ||||||
| Diagonal covariance | ||||||
| Static-factor covariance | ||||||
| Elpac (full) |
Appendix figures & tables15 assets
Supplementary material from the paper’s appendix.
Appendix
| Section | Contents |
|---|---|
| App. B | Full model specification and inference: joint likelihood, variational family, ELBO and reparameterized gradients, training algorithm, test-time inference, and closed-form derived summaries (Sylvester–Woodbury reductions, pairwise correlation, Fisher modality weights) |
| App. C | Formal identifiability definitions, full proof of Theorem 1, warp invariance, the Theorem-1 checklist at the fitted real-cohort posterior, and full-model identification proposition |
| App. D | Experimental protocol: cohort summary, informative priors, competing methods and runtime, metric definitions, sensitivity analysis (hyperparameters and cohort size), robustness to anchor corruption, and test-time refinement |
| App. E | Real-data validation and biology: leave-one-cohort-out cohort-shift validation, longitudinal validation (stage velocity, cognition recovery, MCI-to-AD conversion), subtype biology summaries (stage discriminability, endpoint pathology, cross-modal correlations), per-factor multimodal mixture, and stage-varying Fisher modality weights |
| Class | Theorem-1 condition | Status |
|---|---|---|
| Structural | ; lower-triangular, strictly positive diagonal; | all hold — at nominal |
| centered coefficients, diagonal-positive | hold by parameterization | |
| non-constant activations | holds in all 20 pooled seeds; every subtype carries stage-monotone columns | |
| Regularity | (i) spline-design coverage | full rank in all 240 designs, worst |
| (ii) Anderson–Rubin row deletion | exhaustive — all 17-row sub-blocks of retain full rank 8 | |
| (iii) distinguishable trajectory features | generic under the structured velocity, which places interior features in ; fitted onsets separate across modalities in every subtype |
| Scope | Subtype | min equation 16 gap over | min , | |
|---|---|---|---|---|
| Pooled | EOAD | |||
| LOAD | ||||
| HA | ||||
| Leave-ADNI | EOAD | |||
| LOAD | ||||
| HA |
| ADNI | NACC | Total | |
| Subjects | |||
| Subtype anchors | |||
| EOAD (early-onset Alzheimer’s disease) | |||
| LOAD (late-onset Alzheimer’s disease) | |||
| HA (healthy aging) | |||
| NC startpoint pool † | |||
| Synthetic | Real | ||||
| Method | Device | Time (s) | Time (s) | ||
| SuStaIn [ 6 ] | CPU | ||||
| DEBM [ 7 ] | CPU | ||||
| Static-FA [ 27 , 47 ] | CPU | ||||
| MoPoE-VAE [ 22 ] | CPU | ||||
| XGBoost [ 35 ] | CPU | ||||
| Scenario | Data | Metric | 5% | 10% | 20% |
|---|---|---|---|---|---|
| Mislabel, clinical confusion | Synthetic | Subtype acc | |||
| Stage | |||||
| Real | Subtype BAcc | ||||
| Macro-F1 | |||||
| Held-out LL | |||||
| Mislabel, uniform-random | Synthetic | Subtype acc |
| Adam steps | 1 | 2 | 5 | 10 | 20 | 50 |
|---|---|---|---|---|---|---|
| Synthetic (100 replicates) | ||||||
| Real cohort (20 splits) |