How bartongroup/ank-analysis works
Key Components
a. Input Data
• files/ (raw multiple-sequence alignments, variant tables)
• (Potential future additions: configuration files or additional datasets)
b. Data-Retrieval & Integration Modules
• retrieve_data_allsp.py — fetches and merges UniProt/InterPro annotations
• structure_validation.py — downloads/validates structural mappings
c. Alignment Modules
• new_aligner.py, alignment_editing.py, slaver.py — sequence-alignment functions
d. Analysis Modules
• variant_analysis.py — computes per-position variant enrichment
• contact_maps.py — builds intra-/inter-repeat contact maps
• structural_analysis.py — calculates RSA, secondary-structure, PPI enrichments
e. Orchestration Layer (Notebooks)
• def_data_extraction.ipynb
• def_database_integration.ipynb
• def_upset.ipynb (intersection plots)
• def_alignment.ipynb
• def_variant_analysis.ipynb
• def_structure_validation.ipynb
• def_contact_maps.ipynb
• def_structural_analysis.ipynb
f. Intermediate & Final Outputs
• Generated dataframes, plots, and summary tables (stored alongside notebooks or in dedicated output directories).
Relationships & Data Flow
1. Raw data (files/) → “retrieve_data_allsp” → merged annotation dataframe
2. Merged annotations → “database_integration” notebook → non-redundant sequences
3. Non-redundant sequences → alignment modules (new_aligner, slaver, editing) → final alignment
4. Final alignment → variant_analysis.py → variant enrichment results
5. Alignment + structure identifiers → structure_validation.py → validated PDB mappings
6. Validated structures → contact_maps.py → contact map results
7. Contact maps + PDB structures → structural_analysis.py → RSA, SS, PPI analyses
8. Each step is invoked and visualized in its corresponding notebook.
Architectural Patterns & Principles
• Modular pipeline: each functional area is a separate Python module + orchestrating notebook
• Data-centric flow: clear source → transform → output stages
• Single responsibility: modules focus on one type of processing
• Notebook layer decouples UI/visualization from core logic