Cell Health Data Supplementary Files

This dataset contains supplementary files related to the <em>cell-health-data</em> repository that were too large to upload on GitHub. Each of these files were saved to and loaded from an external hard drive during Cell Health Data Processing. These files include: single_cell_classification_probabilities/ : Single cell classification probabilities derived with each model trained in phenotypic_profiling_model. Each compressed csv file in single_cell_classification_probabilities/ corresponds to single-cell classifications from a particular plate as derived with a particular model. The model used to derive the probabilities is indicated by the file's parent folders. Each compressed csv file includes the metadata (location and perturbation) and phenotypic class probabilities for each cell in the plate. classification_profiles/ : This folder contains aggregated data known as "classification profiles". These profiles are generated from the single-cell classification probabilities found in the 'single_cell_classification_probabilities/' folder. Specifically, for each model, we find the mean of the single-cell classification probabilities across each perturbation and cell line to create a composite profile. This aggregated data provides a summarized view of cell behavior for each perturbation/cell line combination, as predicted by each model. For more information regarding the generation of this data, please see cell-health-data/4.classify-single-cell-phenotypes.本数据集包含与<em>cell-health-data</em>仓库相关的补充文件，此类文件因体积过大无法上传至GitHub平台。在细胞健康数据处理（Cell Health Data Processing）流程中，所有此类文件均通过外置硬盘进行存储与读取。 此类文件包含以下内容： 1. single_cell_classification_probabilities/ 文件夹：存储由表型分析模型（phenotypic_profiling_model）中训练的各模型生成的单细胞分类概率（single cell classification probabilities）。该文件夹下的每个压缩CSV文件，均对应某一模型针对特定微孔板生成的单细胞分类结果。生成该概率所使用的模型，可通过该文件所在的父级文件夹进行识别。每个压缩CSV文件均包含该微孔板内所有细胞的元数据（位置与扰动处理信息）以及表型类别概率。 2. classification_profiles/ 文件夹：该文件夹存储被称为"分类特征集（classification profiles）"的聚合数据。此类特征集由single_cell_classification_probabilities/文件夹内的单细胞分类概率生成。具体而言，针对每一个模型，我们会计算同一扰动处理与细胞系组合下所有单细胞分类概率的均值，以此生成综合特征集。该聚合数据可直观展示各模型预测的、每一组扰动处理/细胞系组合对应的细胞行为特征。如需了解该数据生成的更多细节，请参阅cell-health-data仓库内的4.classify-single-cell-phenotypes文件。