Skip to content

Build an evidence network#

An evidence network helps explain which accepted peptides support which reported proteins. FastaLake exports the complete relationships and can render a bounded view through the Clinical Knowledge Graph (CKG) tools. The renderer is experimental; the complete evidence tables remain available without CKG or a graph database.

flowchart LR
    S[Completed study groups] --> E[Peptide–protein evidence export]
    F[Recorded protein sequences] --> A[Exact sequence annotation join]
    N[Native eggNOG output] --> A
    E --> A
    E --> C[Optional CKG view]
    A --> C
    A --> T[Functional tables and annotation coverage]

Export and add existing eggNOG results#

fasta-lake network export --groups grouped --out evidence_network

fasta-lake network eggnog --bundle evidence_network \
  --annotation annotations \
  --fasta grouped/representatives.fasta --out network_functions

annotations is a completed result from FastaLake annotation. For native eggNOG output, pass its filename as --annotation and add --query-fasta original_annotation_queries.fasta. Keep the original query sequences: matching names alone is insufficient. The selected FASTA must be the recorded study representatives or a recorded search FASTA.

The join checks full protein sequence identity. Different identifiers may share the same exact sequence; I/L-equivalent proteins are still distinct annotation queries. No-hit, unmatched and unavailable sequences remain in protein_annotations.tsv. Protein-to-function edges record annotation, not measured activity. An annotation is never propagated into a claim that an ambiguous peptide came uniquely from that protein.

KO, EC, GO and PFAM are retained by default; --field selects namespaces. group_terms_KEGG_ko.tsv and equivalent files can feed the existing functional exploration command with your selected sum/directLFQ matrix and declared metadata. The biological interpretation guide distinguishes commonly shared functions from specialised community capabilities.

Render with a compatible CKG environment#

Use a separately configured CKG installation. The worker runs in its Python environment and calls its analytics and visualization modules; it does not connect to Neo4j. The compatibility notes describe the tested environment and small proposed upstream fixes.

fasta-lake network render --bundle evidence_network \
  --annotations network_functions \
  --ckg-python /path/to/ckg-environment/bin/python \
  --ckg-source /path/to/compatible-CKG \
  --max-nodes 150 --max-edges 300 --out network_report

Open network_report/index.html. The interactive view, PNG/PDF/SVG overview and full evidence count plot are local files. Hover over a node for identifiers and support. Full node/edge tables remain in the original export.

  • Solid edge: assigned reporting representative.
  • Dashed edge: another reported peptide membership.
  • Purple dotted edge: eggNOG orthology annotation.

The view is intentionally bounded and reports visible versus total counts. KO/EC annotations appear in the overview; the full annotation export retains the additional namespaces. CKG's topological communities describe this graph's connections. They are not pathways, protein interactions or an independent protein inference result.

The offline Docker demo includes a real CAMPI evidence export and one recorded enolase annotation. This demonstrates the join with deliberately partial coverage. CKG rendering uses the separate optional environment above.

Credit and provenance#

Network analysis and interactive plots reuse CKG, NetworkX, Plotly and PyVis. FastaLake preserves the peptide evidence and adds the exact annotation join. RENDER_COMPLETE.json records source hashes, versions, view limits and outputs. The CKG and eggNOG citations appear in reports and exported network figures.

  • Santos A et al. A knowledge graph to interpret clinical proteomics data. Nature Biotechnology 40:692–702 (2022). doi:10.1038/s41587-021-01145-6.
  • Cantalapiedra CP et al. eggNOG-mapper v2: functional annotation, orthology assignments, and domain prediction at the metagenomic scale. Molecular Biology and Evolution 38:5825–5829 (2021). doi:10.1093/molbev/msab293.