#36 · Single-cell

Interpretable models of cell-fate trajectories and spatial gene expression fields

The common thread is making trajectory structure explicit so that outputs can be interpreted. FusedFCR fits a regularized forward continuation-ratio model with lasso and fusion penalties to select stage-specific markers along cell-fate trajectories and projects cells onto a pseudotime-like axis. scCS quantifies fate commitment, speed, affinity and entropy through discounted future-fate propagation, delivered as a scverse-compatible Python framework. Loom applies pseudo-temporal trajectories to spatial transcriptomics, combining cross-sample comparison with local microenvironment analysis in a visual computing system. Spaceland extends spatial profiling to whole-organ scale, reconstructing continuous high-resolution 3D gene-expression fields from sparse spatial transcriptomics and serial H&E histology. Across these items, stage-specific selection, fate metrics, and spatial reconstruction are each made explicit rather than left implicit in embeddings.