FLUX model uncovers biological state transitions from unpaired snapshots
New AI framework handles messy biological data with geometry-aware flow matching.
FLUX (FLow matching for Unpaired longitudinal data with miXture-of-experts) addresses a critical challenge in biological modeling: many systems evolve through continuous dynamics while switching between latent regimes (learning, developmental stages, internal states), yet observations are often unpaired longitudinal snapshots—the same cells or animals are not tracked over time. The framework jointly performs transport modeling and unsupervised regime discovery. It learns a data-dependent metric from pooled labeled and unlabeled observations, constructs geometry-aware conditional paths between adjacent marginals, and decomposes the resulting velocity field into sparse expert vector fields via a Straight-Through Gumbel-Softmax router.
Across manifold controls, a regime-switching Lorenz system, widefield cortical calcium imaging during associative learning, and embryoid body single-cell differentiation, FLUX reconstructs longitudinal transport while recovering interpretable regime structures. Ablation studies confirm that mixture-of-experts routing alone is insufficient: without geometric learning, FLUX fits local transport but fails to discover regimes encoded in local dynamics. This suggests that geometry-aware velocity decomposition provides a general strategy for discovering latent biological state transitions from unpaired longitudinal snapshots.
- FLUX uses a data-dependent metric to handle curved low-dimensional manifolds in high-dimensional biological data
- Employs a Straight-Through Gumbel-Softmax router to select among sparse expert vector fields for different regimes
- Validated on Lorenz system, cortical calcium imaging, and single-cell differentiation; outperforms models without geometric learning
Why It Matters
Enables discovery of hidden biological states from fragmented longitudinal data, advancing neuroscience and developmental biology.