rtemis/depmap-prism-idasanutlin

DepMap gene expression and PRISM idasanutlin response

Gene expression and response to the MDM2 inhibitor idasanutlin in 454 cancer cell lines. Expression is log2(TPM + 1) of the 1,000 protein-coding genes with the largest variance across these cell lines, from DepMap 24Q2 Public. Response is the log2 fold change in viability after 5 days at 2.5 µM, relative to vehicle control, from the PRISM Repurposing 19Q4 secondary screen. The dataset is a single Parquet file with one row per cell line.

Idasanutlin inhibits MDM2, the protein that marks p53 for degradation. Cell lines with functional p53 lose viability; cell lines with inactivated p53 do not.

Structure

data/depmap-prism-idasanutlin.parquet: 454 rows, 1,004 columns. There are no missing values.

Column	Type	Description
cell_line_id	string	DepMap model identifier (ACH-) of the cell line.
cell_line_name	string	Cell line name.
lineage	categorical	Lineage of the cell line in the OncoTree classification; 22 lineages.
CFH … TMEM265	double	log2(TPM + 1) of one protein-coding gene; 1,000 genes, named by HGNC symbol.
idasanutlin_log2fc_2_5um	double	log2 fold change in viability after idasanutlin at 2.5 µM, relative to vehicle (DMSO).

idasanutlin_log2fc_2_5um is the last column. 0 corresponds to no change in viability, and negative values to loss of viability: -1 is half the viability of vehicle control. Its values range from -4.25 to 1.41, with median 0.08. The dataset has one train split that holds all 454 cell lines.

cell_line_id, cell_line_name, and lineage describe the cell lines. For expression-based prediction, use the gene columns as features and keep lineage to examine how predictions vary by lineage.

lineage	Cell lines
Lung	91
Skin, Esophagus/Stomach, CNS/Brain, Pancreas	31 to 37 each
13 further lineages	6 to 26 each
Peripheral Nervous System, Prostate, Ampulla of Vater, Fibroblast	1 to 5 each

Load

R - using rtemis (1.4.3 or later):

dat <- read("hf://datasets/rtemis/depmap-prism-idasanutlin")

R - using croissant, with lineage as a factor:

library(croissant)
dat <- read_record_set(
  "hf://datasets/rtemis/depmap-prism-idasanutlin/croissant.json",
  "cell_lines"
)

R - using nanoparquet:

url <- "https://huggingface.co/datasets/rtemis/depmap-prism-idasanutlin/resolve/main/data/depmap-prism-idasanutlin.parquet"
path <- file.path(tempdir(), "depmap-prism-idasanutlin.parquet")
download.file(url, path, mode = "wb")
dat <- nanoparquet::read_parquet(path)

Python - using polars:

import polars as pl
dat = pl.read_parquet("hf://datasets/rtemis/depmap-prism-idasanutlin/data/depmap-prism-idasanutlin.parquet")

Notes

• The 1,000 genes were selected by variance of expression across the 454 cell lines, without reference to the outcome. They include the p53 target genes PHLDA3 and GDF15.
• The outcome is a single-dose measurement and carries assay noise. The other 7 doses of the screen are not included.
• Expression (DepMap 24Q2) and drug response (PRISM 2019) were measured at different times, in different cultures of each cell line.
• Lineage frequencies are unequal, and several lineages have few cell lines.
• Cell lines are models of cancer and differ from tumors in vivo.

Source and processing

Source files:

• DepMap 24Q2 Public: OmicsExpressionProteinCodingGenesTPMLogp1.csv (expression) and Model.csv (cell line names and OncoTree lineages).
• PRISM Repurposing 19Q4: secondary-screen-replicate-collapsed-logfold-change.csv, secondary-screen-replicate-collapsed-treatment-info.csv, and secondary-screen-cell-line-info.csv.

PRISM measures viability with pooled, barcoded cell lines. Log fold changes are relative to DMSO, corrected for experimental confounders with ComBat, and collapsed across replicates by the median. The column for idasanutlin at 2.5 µM in screen MTS010, which the PRISM documentation recommends where available, was selected for cell lines that passed STR profiling and have DepMap 24Q2 expression. The 1,000 genes with the largest variance of log2(TPM + 1) across these cell lines were kept, in the order of the DepMap file, and named by HGNC symbol. Values are unchanged.

Metadata

croissant.json is the dataset's Croissant 1.1 metadata, written with the croissant R package. It describes the file with its SHA-256 checksum, each field, including the Entrez Gene ID of each gene, and the Responsible AI properties of the data. lineage references the enumeration record set cell_lines_lineage_levels. The Parquet file stores lineage as strings.

License

CC BY 4.0, the license of the DepMap 24Q2 Public and PRISM Repurposing 19Q4 releases.

Citation

Cite the DepMap release, the PRISM release, and the PRISM paper:

@misc{depmap2024q2,
  author = {{Broad DepMap}},
  title = {{DepMap 24Q2 Public}},
  year = {2024},
  howpublished = {figshare},
  doi = {10.6084/m9.figshare.25880521.v1}
}

@misc{prism2019q4,
  author = {Corsello, Steven and Kocak, Mustafa and Golub, Todd},
  title = {{PRISM Repurposing 19Q4 Dataset}},
  year = {2019},
  howpublished = {figshare},
  doi = {10.6084/m9.figshare.9393293.v4}
}

@article{corsello2020discovering,
  author = {Corsello, Steven M. and Nagari, Rohith T. and Spangler, Ryan D. and Rossen, Jordan and Kocak, Mustafa and Bryan, Jordan G. and Humeidi, Ranad and Peck, David and Wu, Xiaoyun and Tang, Andrew A. and Wang, Vickie M. and Bender, Samantha A. and Lemire, Evan and Narayan, Rajiv and Montgomery, Philip and Ben-David, Uri and Garvie, Colin W. and Chen, Yejia and Rees, Matthew G. and Lyons, Nicholas J. and McFarland, James M. and Wong, Bang T. and Wang, Li and Dumont, Nancy and O'Hearn, Patrick J. and Stefan, Eric and Doench, John G. and Harrington, Caitlin N. and Greulich, Heidi and Meyerson, Matthew and Vazquez, Francisca and Subramanian, Aravind and Roth, Jennifer A. and Bittker, Joshua A. and Boehm, Jesse S. and Mader, Christopher C. and Tsherniak, Aviad and Golub, Todd R.},
  title = {Discovering the anticancer potential of non-oncology drugs by systematic viability profiling},
  journal = {Nature Cancer},
  year = {2020},
  volume = {1},
  number = {2},
  pages = {235--248},
  doi = {10.1038/s43018-019-0018-6}
}