CCLE_expression.ipynb

CCLE_expression.ipynb — Gene-expression preprocessing

Objective

Extract a cancer-focused subset of the raw CCLE 22Q2 expression matrix and give it clean gene-symbol headers for downstream joins.

---

Inputs

File	Drive location	Notes
CCLE_expression.csv	CCLE-DepMap22Q2_geneexp/	1 ,406 cell lines × 19 ,221 genes (log₂(TPM + 1))
CancerGeneCensus_GRCh38_COSMIC_v101.csv	same folder	List of 735 cancer-associated genes
sample_info.csv	same folder	CCLE sample ↔ DepMap ID mapping (for later alignment)

---

Processing steps ↴

1. Load full matrix with pandas.read_csv (mixed-dtype header fix).

2. Subset to COSMIC genes → 735 columns retained.

3. Rename columns from Ensembl IDs to HGNC gene symbols (e.g., ENSG00000141510 → TP53).

4. Integrity checks:

   • No missing values per gene.

   • 1 ,406 unique DepMap IDs confirmed.

5. Save cleaned table as CCLE_expression_cleaned.csv (≈ 16 MB).

---

Outputs consumed later

File	Used by
CCLE_expression_cleaned.csv	01_data_prep.ipynb

---

Rationale

• Focusing on 735 cancer-associated genes reduces dimensionality 20-fold while preserving biologically relevant signals.

• Leaving expression on a log₂(TPM + 1) scale avoids negative values and allows z-scoring later in Data Integration & Preparation.ipynb.

• A lightweight 16 MB CSV is quick to load in Colab and small enough to version-control if desired.

No additional files are written; intermediate DataFrames stay in memory only.