Predicting cell morphological responses to perturbations using generative modeling

Pre-processed datasets and checkpoints for the paper "Predicting cell morphological responses to perturbations using generative modeling". Datasets BBBC021 six: the BBBC021 dataset pre-processed and downsampled to 6 drugs. Images are cropped around single cells. BBBC021 all: the BBBC021 dataset pre-processed for images from 99 drug treatments. A larger field of view is considered here. cpg0000: a subset of the JUMP cpg0000 dataset pre-processed for the U2OS cell lines. cpg0000 transported: a subset of the JUMP cpg0000 dataset pre-processed for the U2OS cell lines and plate-corrected with IMPA. RxRx1: the RxRx1 dataset subset for U2OS cells and 3 batches. Checkpoints For each dataset, checkpoints for the trained IMPA model are provided. For the BBBC021 all, BBBC021 six and the cpg0000 transported datasets, the checkpoints are from a model trained to perform perturbation prediction. On the untrasported cpg0000 and RxRx1 the checkpoints are for batch correction.本资源为论文《基于生成式建模预测扰动下细胞形态响应》（Predicting cell morphological responses to perturbations using generative modeling）配套的预处理数据集与模型检查点（checkpoint）。 ## 数据集 BBBC021 six：经预处理并筛选至6种药物的BBBC021数据集，所有图像均围绕单个细胞进行裁剪。 BBBC021 all：针对99种药物处理图像完成预处理的BBBC021数据集，本版本采用更大的视野范围。 cpg0000：JUMP cpg0000数据集的子集，针对U2OS细胞系完成预处理。 cpg0000 transported：JUMP cpg0000数据集的子集，针对U2OS细胞系完成预处理，并通过IMPA完成板间校正。 RxRx1：针对U2OS细胞与3个批次的RxRx1数据集子集。 ## 模型检查点 所有数据集均配套提供训练完成的IMPA模型的检查点。其中，针对BBBC021 all、BBBC021 six与cpg0000 transported数据集的检查点，来自用于执行扰动预测任务的训练模型；而针对未完成板校正的cpg0000与RxRx1数据集，其检查点用于批次校正任务。