recursionpharma/rxrx3-core

To accompany OpenPhenom, Recursion is releasing the RxRx3-core dataset, a challenge dataset in phenomics optimized for the research community. RxRx3-core includes labeled images of 735 genetic knockouts and 1,674 small-molecule perturbations drawn from the RxRx3 dataset, image embeddings computed with OpenPhenom, MAE-L/8, MAE-G/8, and associations between the included small molecules and genes. The dataset contains 6-channel Cell Painting images and associated embeddings from 222,601 wells but is less than 18Gb, making it incredibly accessible to the research community.

Mapping the mechanisms by which drugs exert their actions is an important challenge in advancing the use of high-dimensional biological data like phenomics. We are excited to release the first dataset of this scale probing concentration-response along with a benchmark and model to enable the research community to rapidly advance this space.

Paper published at LMRL Workshop at ICLR 2025 RxRx3-core: Benchmarking drug-target interactions in High-Content Microscopy.
Benchmarking code for this dataset is provided in the EFAAR benchmarking repo and Polaris.

---

Loading the RxRx3-core image dataset

from datasets import load_dataset
rxrx3_core = load_dataset("recursionpharma/rxrx3-core")

Loading OpenPhenom embeddings and metadata for RxRx3-core

from huggingface_hub import hf_hub_download
import pandas as pd

file_path_metadata = hf_hub_download("recursionpharma/rxrx3-core", filename="metadata_rxrx3_core.csv",repo_type="dataset")
file_path_embs = hf_hub_download("recursionpharma/rxrx3-core", filename="OpenPhenom_rxrx3_core_embeddings.parquet",repo_type="dataset")

open_phenom_embeddings = pd.read_parquet(file_path_embs)
rxrx3_core_metadata = pd.read_csv(file_path_metadata)

---

Metadata

The metadata can be found in metadata_rxrx3_core.csv in this repository. The schema of the metadata is as follows:

Attribute	Description
well_id	Experiment Name - Plate - Well (compound-004_1_AA04 or gene-088_9_C15)
experiment_name	Experiment Name: Experiment number (compound-004 or gene-088)
plate	Plate number in the experiment (1-48)
address	Well location on the plate - "A01" to "AF48".
gene	Unblinded or anonymized gene name, or a control
treatment	Compound synonym or gene-name - guide-number (Narlaprevir or _guide_1)
SMILES	Canonical SMILES or blank for non-compounds
concentration	Compound concentration tested (in uM)
perturbation_type	CRISPR or COMPOUND
cell_type	HUVEC
well_type_label	Indicates experimental control information

The well_type_label column includes the following values:

well_type_label	Description
Query guides	CRISPR guides that target a query gene
Exon controls	Exon-targeting CRISPR guides that are used as controls
Intron controls	Intron-targeting CRISPR guides that are used as controls
Query Compounds + Intron control	Query compounds on an intron-targeting CRISPR background
CRISPR Gene Positive Controls	Control genes that are exon-targeting CRISPR guides that are used as controls, there are five genes with 6 guides each that target the exon region of the gene
Control Compounds + Intron control	Control compound on an intron-targeting CRISPR background

To help understand the metadata, we have included some samples to enable parser testing and validation

well_id,experiment_name,plate,address,gene,treatment,SMILES,concentration,perturbation_type,cell_type,well_type_label
compound-001_10_AA12,compound-001,10,AA12,,Esomeprazole,"COC1=CC2=C([N-]C(=N2)[S@@](=O)CC2=C(C)C(OC)=C(C)C=N2)C=C1 |r,c:7,13,21,24,t:2,4,18|",2.5,COMPOUND,HUVEC,Control Compounds + Intron control
gene-077_3_L32,gene-077,3,L32,CENPC,CENPC_guide_3,,,CRISPR,HUVEC,Query guides