Configuration
CAPELLINI is driven by a single YAML configuration file that you provide.
The package does not ship any bundled configs: it simply remembers the
path of the last config you loaded — stored at
~/.capellini/last_config — and re-uses it on the next run.
In the terminal UI, point CAPELLINI at your YAML via Settings → Load config. Programmatic users pass the path directly:
from capellini import CapelliniConfig, CapelliniPipeline
cfg = CapelliniConfig.from_yaml("/path/to/your_config.yaml")
CapelliniPipeline(cfg).run_all()
Key parameters
Parameter |
Default |
Meaning |
|---|---|---|
|
|
Genus-level ( |
|
|
Rank used to aggregate bacteria for the network stage |
|
|
Keep features present in ≥ 10 % of samples |
|
|
Strength of taxonomy smoothing applied to |
|
|
Strength of CRISPR-informed abundance propagation |
|
|
Number of message-passing updates |
|
|
SpacePHARER FDR threshold |
|
|
MinCED minimum spacers per array |
Inputs
CAPELLINI expects, per cohort:
Raw 16S rRNA amplicon FASTQ files (forward, reverse, or paired).
A viral contig FASTA (e.g. ViroProfiler output).
A sample metadata CSV used to align bacterial and viral abundances.
The SILVA reference (Release 138.1) and SILVA taxmap.
All output folders under base/ are created automatically.
Outputs
For each study, CAPELLINI writes under Enhanced Networks/<study>/:
common/— aligned, prevalence-filteredV,B, and metadata tables.shrinkage/— Schäfer–Strimmer shrinkage correlations on the CLR-stacked \(Z = [B^{\mathrm{CLR}}\ V^{\mathrm{CLR}}]\).crispr_raw/andcrispr_smooth/— raw and taxonomy-smoothed CRISPR matrices (\(W\), \(\tilde{W}\)).xstar/— host-informed abundances \(Z^*\) from convex and residual message-passing variants.