recursionpharma/rxrx3-core

RxRx3-core数据集是一个为研究社区优化的表型组学挑战数据集，伴随OpenPhenom发布。它包括735个基因敲除和1,674个小分子扰动的标记图像，使用OpenPhenom计算的图像嵌入，以及包含的小分子与基因之间的关联。数据集包含222,601个孔的6通道Cell Painting图像和相关嵌入，大小小于18Gb，便于研究社区访问。该数据集在药物作用机制映射中具有重要作用，并提供了加载数据集和嵌入的代码示例。

RxRx3-core is a challenge dataset in phenomics optimized for the research community, including labeled images of 735 genetic knockouts and 1,674 small-molecule perturbations, image embeddings computed with OpenPhenom, and associations between the included small molecules and genes. The dataset contains 6-channel Cell Painting images and associated embeddings from 222,601 wells, totaling less than 18Gb, making it highly accessible to the research community.