Skip to content

CLAUDE.md

Latest commit

 

History

History
62 lines (50 loc) · 3.43 KB

CLAUDE.md

File metadata and controls

62 lines (50 loc) · 3.43 KB
Raw

CLAUDE.md

Project Overview

This is a COSMOS (Causal Oriented Search of Multi-Omic Space) tutorial/analysis project using the NCI60 cancer cell line panel. It demonstrates multi-omics data integration by connecting transcription factor activities (upstream) to metabolomic measurements (downstream) through a prior knowledge network (PKN), using the cosmosR R package and the MOON (Multi-Omics Optimized Network) scoring algorithm.

Language & Framework

  • Language: R
  • Core package: cosmosR (installed from GitHub: saezlab/cosmosR)
  • Key dependencies: decoupleR, reshape2, readr, ggplot2, GSEABase, igraph
  • Project type: RStudio project (NCI60.Rproj)
  • Main analysis: R Markdown (net_compr_MOON.Rmd)

Directory Structure

├── data/                  # Input data
│   ├── cosmos/            # Pre-processed COSMOS inputs (RData, per cell line)
│   ├── metabolomic/       # Raw metabolomic data
│   └── RNA/               # RNA expression data
├── scripts/               # R scripts
│   ├── prepare_cosmos_inputs.R   # Data preprocessing
│   ├── net_compr_MOON.R          # Main MOON analysis (script version)
│   ├── net_compr_carnival.R      # CARNIVAL-based analysis (compressed)
│   ├── net_no_compr_carnival.R   # CARNIVAL-based analysis (uncompressed)
│   ├── RNA.R / metabolomic.R     # Data processing scripts
│   ├── support_compression.R     # Network compression utilities
│   ├── support_decompression.R   # Network decompression utilities
│   └── support_decoupleRnival.R  # decoupleR/CARNIVAL support functions
├── support/               # Support files (regulons, pathway GMTs, helper functions)
├── results/               # Output results (SIF/ATT network files, MOON scores)
│   └── moon/              # MOON scoring results per cell line
├── EBI_practical/         # Presentation materials
├── cbc/                   # CBC tools
├── net_compr_MOON.Rmd     # Main R Markdown tutorial
└── net_compr_MOON_files/  # Rendered figures from Rmd

Key Concepts & Workflow

  1. Load data: TF activity scores, metabolomics, RNA expression for a given NCI60 cell line
  2. Prepare inputs: Assign metabolite compartments, filter significant TF inputs
  3. Filter PKN: Remove unexpressed genes, prune network to reachable/observable nodes within n_steps (default: 6)
  4. Compress network: compress_same_children() to reduce redundant parent nodes
  5. Run MOON: Iterative scoring with moon() + filter_incohrent_TF_target() until convergence
  6. Decompress & extract: Decompress results, apply cutoff, extract solution subnetwork (SIF + ATT format)
  7. Pathway enrichment: Optional Reactome pathway scoring via decoupleR::run_ulm()

Output Formats

  • SIF: Source-Interaction-Target edge list (columns: source, target, sign)
  • ATT: Node attribute table (columns: Nodes, score, plus activity values)
  • Results are written as CSVs to results/

Conventions

  • Cell line names follow NCI60 naming (e.g., "786-0", "A549/ATCC", "MCF7")
  • Metabolite IDs use HMDB format with compartment suffix (e.g., Metab__HMDB0011747_c)
  • Internal cosmosR functions are accessed via cosmosR::: (triple colon)
  • Network operations use data frames with source/target/interaction columns
  • DoRothEA regulons are loaded from support/dorothea_reg.RData