Skip to content

CLAUDE.md

Latest commit

 

History

History
59 lines (43 loc) · 3.35 KB

CLAUDE.md

File metadata and controls

59 lines (43 loc) · 3.35 KB
Raw

CLAUDE.md — Agent Rules for MAVEable-Genes

This file provides persistent guidance for AI agents working in this repository.

Project Overview

This repository filters ClinVar variant data by gene attributes to identify genes amenable to Multiplexed Assays of Variant Effect (MAVEs). The primary workflow is a Jupyter notebook (ClinVar_wrangle_uniprot_gencc_merge.ipynb) that produces a ranked list of MAVEable genes based on criteria such as essentiality, subcellular localization, protein complex membership, and known pathogenic variant associations.

Key Files

  • ClinVar_wrangle_uniprot_gencc_merge.ipynb — main analysis notebook; do not restructure without understanding the full pipeline
  • MAVEable_final_is_feasible.csv — final output with feasibility flags; treat as a derived artifact
  • unique_gencc.csv — GenCC gene-disease associations (moderate confidence and above)
  • uniprot_len_loc_2.csv — UniProt protein length and localization data
  • subcellular_location.tsv — subcellular localization reference
  • hap1_essential.csv.xlsx — HAP1 cell essentiality data
  • Tcell_screen.csv — T cell essentiality screen
  • VAMPseq-ome.csv — VAMPseq toxicity data
  • opencell-library-metadata.csv — OpenCell complex/interaction data
  • Rendo_2025_toxic.csv — exogenous expression toxicity data
  • lacoste.csv — additional reference dataset
  • instance_order.ipynb — secondary notebook; purpose should be confirmed before editing

The large ClinVar source file (variant_summary_*.txt.gz) is not committed to this repo and must be downloaded from the NCBI FTP site.

Language & Libraries

  • Python 3 with Jupyter notebooks exclusively
  • Core libraries: pandas, numpy, matplotlib
  • Do not introduce new dependencies without confirming with the user

Coding Conventions

  • Use pandas for all data manipulation; avoid raw Python loops over DataFrames where vectorized operations are available
  • Filter steps should be explicit and commented with the biological rationale (e.g., why a ClinicalSignificance value is excluded)
  • Preserve existing variable naming conventions (e.g., ClinVar_01_25_SNV_simple, unique_gencc)
  • When adding new filter steps, append them in the logical order of the existing pipeline rather than inserting mid-cell
  • Use case=False, na=False in .str.contains() calls for robustness
  • Three-letter amino acid codes should be converted to single-letter using the existing amino_acid_map dictionary pattern

Data Handling

  • Treat all CSV/TSV files as read-only reference data unless the user explicitly asks to modify them
  • When reading large files (e.g., ClinVar), use low_memory=False or explicit dtype arguments to avoid mixed-type warnings
  • Do not hard-code absolute file paths; assume files are in the working directory
  • The ClinVar assembly should always be filtered to GRCh38

Output Expectations

  • Final outputs are CSV files prefixed with MAVEable_
  • Plots use matplotlib; maintain existing plot style unless a change is requested
  • Do not overwrite existing output CSVs without explicit user confirmation

What Not to Do

  • Do not delete or rename existing CSV/TSV data files
  • Do not reorder or consolidate notebook cells without understanding pipeline dependencies
  • Do not commit the large ClinVar .txt.gz source file
  • Do not add unrelated dependencies or change the Python environment