References and software context

scCS is designed to operate on AnnData objects and to complement established single-cell analysis and RNA-velocity workflows.

Core software ecosystem

scCS interoperates with:

• AnnData for annotated matrix storage;

• Scanpy for preprocessing, neighborhood graphs, dimensionality reduction, clustering, and visualization;

• scVelo for RNA-velocity estimation, velocity graphs, latent time, and velocity pseudotime;

• RegVelo datasets and models where available;

• SciPy, NumPy, pandas, statsmodels, and related scientific Python tools.

Scientific scope

RNA velocity, latent time, velocity pseudotime, source transition graphs, and terminal-state evidence remain responsibilities of the corresponding upstream methods. scCS adds supervised furcation-specific quantification, condition comparison, interpretation, downstream analysis, and standardized visualization.

Methodological context

The methodological tutorials compare DFFP with immediate transition pushforward, scVelo embedding projection, graph refitting on supervised coordinates, local trajectory-frame projection, first-exit absorption, and unlimited absorption. These comparisons explain the final method choice and are not required steps in routine user analyses.

For the equations, assumptions, and interpretation of scCS outputs, see:

• Mathematical framework

• Metrics and semantics

• Method selection

• Limitations