# Open Problems NeurIPS 2023 Single-Cell Perturbations: 146 Compounds in Human PBMCs

## Facts

- Provider: Open Problems in Single-Cell Analysis (Szałata, Cannoodt, Burkhardt, Luecken et al.)
- License: CC-BY-4.0 (https://proceedings.neurips.cc/paper_files/paper/2024/file/24c4d51f3ef48dd2dbab78243ecb26a1-Paper-Datasets_and_Benchmarks_Track.pdf)
- Stored at: Amazon S3, s3://openproblems-bio/public/neurips-2023-competition (us-east-2)
- Access: public, readable without an account
- Category: Drug Discovery
- Source: https://openreview.net/forum?id=WTI4RJYSVm
- Page: https://scigantic.com/archive/ccd6368a-e656-5fc9-bdee-fbede800b471
- Keywords: Drug Discovery, single-cell RNA-seq, AWS Open Data

## About

Single-cell RNA-seq of human peripheral blood mononuclear cells from three healthy donors, each treated for 24 hours with 146 small molecules chosen from the LINCS Connectivity Map library, generated for the NeurIPS 2023 perturbation prediction competition. Compounds were dosed at 1 µM (belinostat 0.1 µM) in 96-well plates, with a DMSO negative control and two positive controls in every pooled row, giving 576 multiplexed samples and 301,785 cells across T, B, NK and myeloid cell types after quality control. The folder holds raw counts (sc_counts.h5ad, with obs fields for compound, dose, timepoint, plate, well, donor and cell type), processed and reannotated counts, pseudobulk matrices, limma differential expression per plate and per cell type, mechanism-of-action annotations (moa_annotations.csv), baseline 10x Multiome RNA and ATAC profiles for each donor (h5mu), and the original Kaggle upload files. Read h5ad files with anndata or scanpy and h5mu files with mudata or muon. Processed counts are also deposited in GEO as GSE279945.

## Load the data

```python
import boto3
from botocore import UNSIGNED
from botocore.client import Config
s3 = boto3.client('s3', region_name='us-east-2', config=Config(signature_version=UNSIGNED))
s3.list_objects_v2(Bucket='openproblems-bio', Prefix='public/neurips-2023-competition/', Delimiter='/')
```

Lists the files with anonymous access. Runs on any machine, with no cloud account.

## Data structure

Example file: `public/neurips-2023-competition/2023-09-14_kaggle_upload/2023-08-31_sc_multiome_expression_atac.h5ad`

## Open it

- In a hosted notebook with the data mounted: https://scigantic.com/archive/ccd6368a-e656-5fc9-bdee-fbede800b471
- Full record as JSON (schema card included): https://api.scigantic.com/api/archives/ccd6368a-e656-5fc9-bdee-fbede800b471
- Over MCP: `get_schema_card` with id `ccd6368a-e656-5fc9-bdee-fbede800b471` at https://api.scigantic.com/mcp

## Related topics

- [Drug Discovery datasets](https://scigantic.com/topics/drug-discovery)
- [Single-cell RNA-seq datasets](https://scigantic.com/topics/single-cell-rna-seq)
