Pipelines

Every analysis step you run from a workflow page maps to one of these pipelines. This reference lists what each one takes as input, what you can configure, and what it produces.

Cell-cell communication

Pipeline	Input	Outputs
Cell-cell communication	Per-cell expression CSV + metadata CSV	TME-annotated interaction scores, pathway aggregation, network, ligand activity, condition comparison, spatial boundary priorities

You can set: expression and metadata files, the condition column and two comparison values, and the permutation count.

Reference algorithms: CellPhoneDB, NicheNet, CellChat.

Ligand activity IO hypotheses

Pipeline	Input	Outputs
Ligand activity	Per-cell expression CSV + metadata CSV (+ optional ligand-target priors)	Ranked ligand activities (AUPR-corrected, AUROC, Pearson, permutation p/q-value), active ligand-target and ligand-receptor links, a prioritized sender→receiver table, TME sender/receiver classes, clinical metadata summaries, tumor-board highlights

You can set: expression and metadata files, target gene presets or a custom gene list, sender/receiver cell types, and comparison condition.

This workflow calls the shared Immunology ligand activity engine — a Python-native port of the NicheNet method (Browaeys et al., Nat Methods 2019) that scores each ligand by how well its regulatory-potential targets predict the receiver's observed gene program, using NicheNet's own prior model as the knowledge base — then adds oncology target presets for exhaustion, interferon response, antigen presentation, suppression, and cytotoxicity. When the hosted prior model isn't configured, the pipeline falls back to a small built-in demo prior and flags results accordingly.

Immuno-oncology profiling

Pipeline	Input	Outputs
Immune profile	Bulk expression CSV (+ optional repertoire CSV)	Deconvolution, TIDE phenotype, exhaustion scores, repertoire diversity, IO response prediction

You can set: expression file, optional repertoire file, therapy context, and tumor type.

Reference algorithms: CIBERSORTx, xCell, TIDE, scirpy.

Mutation landscape

Pipeline	Input	Outputs
Mutation landscape	MAF-style CSV (+ optional trinucleotide context, copy number/GISTIC-like peak, and clinical CSVs)	TMB, oncoprint, SBS-6/SBS-96 signature exposure, cosine-similarity assignment, co-occurrence/exclusivity, focal/arm CNA summaries, known driver CNA annotations, pathway enrichment, differential mutation testing, clinical CNA/pathway enrichment

You can set: mutation file, optional copy number and clinical files, the clinical grouping column, panel size, and number of top genes shown.

Reference algorithms: maftools, SigProfilerPy.

Malignant cell detection

Pipeline	Input	Outputs
Malignant cell detection	CompBio single-cell CNV handoff/result (+ optional mutation, communication, or survival results)	Malignant populations, CNV evidence, subclone TME features, mutation overlap, survival/response feature export, workflow handoffs, caveats

This pipeline reuses the Computational Biology single-cell CNV engine output. Oncology does not duplicate copy-number inference; it turns the handoff into tumor biology summaries that can feed communication, immune, mutation, survival, and interpretation workflows.

Survival analysis

Pipeline	Input	Outputs
Survival	Clinical CSV (+ optional feature and longitudinal CSVs)	Kaplan-Meier curves, log-rank test, Cox regression, longitudinal trajectories

You can set: clinical file, optional feature/longitudinal files, the stratification field, and Cox covariates.

Cohort comparison

Pipeline	Input	Outputs
Cohort comparison	Completed oncology source runs	Ranked integrated evidence table across communication, immune/TME, mutation, CNA, signature, and survival outputs

If you don't pick specific source runs, the latest completed run of each type for the study is used.

The comparison layer reuses statistics already computed by source pipelines and adds multiple-testing correction when source rows provide p-values. Descriptive rows without p-values are ranked by effect size.

Interpretation

Pipeline	Input	Outputs
Interpret	Completed source runs you select	Narrative with metric citations and research disclaimer

Shared with Computational Biology

These capabilities are not duplicated in oncology pipelines:

Capability	Owner
Single-cell QC, clustering, DE	Computational Biology
Spatial transcriptomics ingestion	Computational Biology
Biomarker ML and pathway enrichment	Computational Biology

Oncology TME views call into compbio infrastructure. See Tumor microenvironment.

Job lifecycle

Every run moves through the same states, shown as a status badge on workflow pages:

queued → running → complete | failed

Versioning

Every run records a pipeline version. Include run IDs and versions in methods sections for reproducibility.

Implementation sources

Oncology-specific algorithms are ported from published reference implementations for each method named above (CellPhoneDB, NicheNet, CellChat, CIBERSORTx, xCell, TIDE, scirpy, maftools, SigProfiler, lifelines).