Informed, scalable and unified transport couplings for single-cell alignment and downstream inference with scBIOT

scBIOT introduces a biologically informed optimal-transport coupling that aligns single cells while preserving known identities, scaling prototype transport to approximately 100 million perturbed cells.

• Why it matters: Single-cell alignment methods often discard the fitted correspondence needed to connect atlas-scale mapping with downstream inference beyond visualization.
• What they did: scBIOT formulates alignment as an explicit, inspectable coupling, uses biological priors to preserve cell identities, compresses transport via prototypes, and derives gene-level changes from coupling-derived transport scores.
• The result: Across transcriptomic and spatial mapping benchmarks, scBIOT balances biological conservation, batch correction, and label-transfer accuracy, enabling state mapping, trajectory inference, perturbation analysis, multimodal comparison, and spatiotemporal localization.