Data archive for leakage-controlled aquatic toxicity prediction

This dataset supports the manuscript "Leakage-Controlled Machine Learning Defines Chemical and Species Extrapolation Limits in Aquatic Toxicity Prediction". It contains the processed aquatic toxicity modeling table, leakage-controlled split assignments, RDKit chemical features, model-performance tables, prediction outputs, error-slice summaries, uncertainty summaries, strong-baseline validation outputs, reference-source holdout outputs, and source-data tables for key manuscript and supplementary figures.

本数据集支撑题为《泄漏控制机器学习（Leakage-Controlled Machine Learning）界定水生毒性预测中的化学物与物种外推极限》的学术手稿。数据集包含经预处理的水生毒性建模表格、泄漏控制拆分分配方案、RDKit化学特征、模型性能表格、预测输出结果、误差切片汇总、不确定性汇总、强基线验证输出结果、参考源预留输出结果，以及用于核心手稿及补充附图的源数据表格。