capellini.stages.network

Network stage: common abundance, shrinkage, raw/smoothed CRISPR, residual X*.

Slim core. The math follows the paper (CLR + W̃ taxonomy smoothing + residual message passing). Schäfer-Strimmer shrinkage correlations are computed both on the joint Z = [B_clr V_clr] and on the propagated Z* = [B_star V_star].

Outputs go directly under cfg.output_root — no per-study sub-folder.

Functions

`build_common_abundance_one`(cfg)
`build_raw_crispr_one`(cfg)	Aggregate the SpacePHARER predictions into a (bac × vOTU) CRISPR matrix.
`build_shrinkage_one`(cfg)	SS shrinkage correlation on Z = [B_clr V_clr].
`build_smooth_crispr_one`(cfg)	Smooth the raw CRISPR matrix via taxonomy kernels; persist all artefacts.
`build_xstar_one`(cfg)	Residual X* propagation + Schäfer-Strimmer correlations on Z*.
`run_network`(cfg)	Run the network stage according to the run_* flags on the config.

capellini.stages.network.build_common_abundance_one(cfg: CapelliniConfig) → dict[str, Path][source]

capellini.stages.network.build_raw_crispr_one(cfg: CapelliniConfig) → Path[source]: Aggregate the SpacePHARER predictions into a (bac × vOTU) CRISPR matrix.

capellini.stages.network.build_shrinkage_one(cfg: CapelliniConfig) → dict[str, Path][source]: SS shrinkage correlation on Z = [B_clr V_clr].

capellini.stages.network.build_smooth_crispr_one(cfg: CapelliniConfig) → dict[str, Path][source]: Smooth the raw CRISPR matrix via taxonomy kernels; persist all artefacts.

capellini.stages.network.build_xstar_one(cfg: CapelliniConfig) → dict[str, Path][source]: Residual X* propagation + Schäfer-Strimmer correlations on Z*.

capellini.stages.network.run_network(cfg: CapelliniConfig) → dict[str, dict][source]: Run the network stage according to the run_* flags on the config.