Published May 12, 2026
| Version v4
Dataset
Open
Phenotypically active ORF, CRISPR and compound consensus profiles
Description
This hosts the files of phenotypically active perturbations for JUMP, derived from https://zenodo.org/records/14025602 (ORF/CRISPR) and from the JUMP COMPOUND profiles (https://github.com/jump-cellpainting/datasets/blob/main/manifests/profile_index.json, subset 'compound') after filtering samples with no cosine-similarity matches above a given threshold (see filenames for thresholds). Two compound variants are provided: compound_filtered_thresh_0_5_p_0_001.gct (~89 MB; corrected p-value <= 0.001) and compound_filtered_thresh_0_5_p_0_2.gct (~289 MB; corrected p-value <= 0.2). Both are derived from profiles_var_mad_int_featselect_harmony.parquet by computing median consensus profiles per JCP2022 id, keeping perturbations with at least one cosine similarity > 0.5 to another perturbation, then restricting to perturbations whose Corrected p-value passes the stated cutoff.