tcga-depmap-drug-profiler/config.yaml at main · Sung2021/tcga-depmap-drug-profiler · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
# ==============================================================================
# TCGA-DepMap Drug Profiler — Configuration
# ==============================================================================

# --- Paths -------------------------------------------------------------------
paths:
  rds_dir: "rds"
  plot_dir: "plots"
  output_dir: "outputs"
  figure_dir: "figures"
  cvae_input_dir: "outputs/CVAE_input"

# --- TCGA-LUAD Processing (01_tcga_data_processing.R) ------------------------
tcga:
  project: "TCGA-LUAD"
  min_sample_pct: 0.5            # Gene must be expressed in ≥50% samples
  tpm_threshold: 1               # Minimum TPM to count as expressed
  mad_quantile_remove: 0.1       # Remove bottom 10% MAD genes

# --- Feature Selection (02_feature_selection.R) ------------------------------
feature_selection:
  n_anova_genes: 3000            # Top ANOVA genes across stages
  n_extreme_genes: 500           # Top Stage I vs IV genes (confounder-adjusted)

# --- DepMap Preparation (03_depmap_preparation.R) ----------------------------
depmap:
  tpm_file: "data/OmicsExpressionTPMLogp1HumanProteinCodingGenesStranded.csv"
  meta_file: "data/Model.csv"
  drug_file: "data/depmap_metadata_drug.csv"
  use_experiment_hub: false      # true = download via ExperimentHub; false = local CSV
  min_tpm: 1                     # Minimum mean TPM for expression filter
  n_hvg: 5000                    # Number of highly variable genes

  # Download URLs for python/download_data.py -----------------------------------
  # Paste direct download links from: https://depmap.org/portal/download/
  # Leave empty ("") to print instructions only; the pipeline will still run
  # if files are already present in data/.
  # MD5 checksums are optional — leave empty ("") to skip verification.
  download_urls:
    tpm_file:      ""              # OmicsExpressionTPM*.csv
    tpm_file_md5:  ""              # optional MD5 checksum
    meta_file:     ""              # Model.csv
    meta_file_md5: ""
    drug_file:     ""              # depmap_metadata_drug.csv
    drug_file_md5: ""

# --- Similarity Analysis (04_similarity_analysis.R) --------------------------
similarity:
  weights:
    global: 0.2                  # S_global  — Spearman on full expression
    pathway: 0.3                 # S_pathway — Pearson on GSVA pathway scores
    target: 0.5                  # S_target  — Cosine on drug-target gene sets
  normalization: "minmax"        # "minmax" (→ [0,1]) or "zscore"

# --- Drug Sensitivity (05_drug_response.R) -----------------------------------
drug_sensitivity:
  method: "fixed"                # "fixed", "percentile", or "stat"
  threshold: -1                  # For method=fixed
  percentile: 25                 # For method=percentile
  k: 1                           # For method=stat (mean - k*sd)
  top_n: 50                      # Top drugs per cell-line report

# --- CVAE (python/cvae_train.py) ---------------------------------------------
cvae:
  latent_dim: 50
  epochs: 300
  batch_size: 32
  learning_rate: 0.001
  beta: 1.0                     # KL divergence weight
  n_per_stage: 100               # Virtual patients per stage (total = 4 × 100)

# --- Network Analysis (06_network_analysis.R) --------------------------------
network:
  similarity_threshold: 0.8      # Edge threshold for network
  max_edges: 500                 # Max edges for visualization