bacbench-antibiotic-resistance-protein-sequences
收藏资源简介:
本数据集包含25,032个细菌基因组,涵盖39个物种,并带有抗微生物药物耐药性标签。基因组蛋白序列是从GenBank中提取的,每个条目包含一个完整的细菌基因组,基因组中的不同contigs用空格分隔。耐药性标签来自Antibiotic Susceptibility Test (AST) Browser,包括二分类(耐药/敏感)标签以及最小抑菌浓度(MIC)的回归值,MIC值经过`log1p`标准化处理。数据集中排除了样本量少的抗生素,最终得到56种抗生素用于MIC回归预测,36种用于二分类标签。对于二分类情况,只包括AST Browser提供的敏感和耐药标签,排除了模糊标签。抗生素的组合被视为单独的药物。标签以CSV文件形式提供,包括二分类标签文件`binary_labels.csv`和MIC回归标签文件`mic_regression_labels.csv`。
This dataset contains 25,032 bacterial genomes spanning 39 species, paired with antimicrobial resistance (AMR) labels. Genomic protein sequences were extracted from GenBank, with each entry holding a complete bacterial genome, where distinct contigs within the genome are separated by spaces. The resistance labels are sourced from the Antibiotic Susceptibility Test (AST) Browser, encompassing both binary (resistant/susceptible) classification labels and regression values of the minimum inhibitory concentration (MIC). The MIC values were normalized using the `log1p` transformation. Antibiotics with inadequate sample sizes were excluded from the dataset. Eventually, 56 antibiotics were retained for MIC regression prediction, while 36 were utilized for binary classification labeling. For the binary classification task, only the susceptible and resistant labels provided by the AST Browser were included, with ambiguous labels excluded. Antibiotic combinations are treated as individual drugs. The labels are provided in CSV format, including the binary classification label file `binary_labels.csv` and the MIC regression label file `mic_regression_labels.csv`.
抗生素耐药性预测数据集概述
数据集基本信息
- 名称: bacbench-antibiotic-resistance-protein-sequences
- 数据来源:
- 基因组蛋白序列: GenBank
- 抗生素耐药性标签: Antibiotic Susceptibility Test (AST) Browser (访问日期: 2024年10月23日)
- 数据量: 25,032个细菌基因组,涵盖39个物种
- 数据类型: 蛋白质序列及抗生素耐药性标签
- 许可证: Apache-2.0
- 标签类型:
- 二元标签 (耐药/敏感)
- 最小抑制浓度 (MIC) 回归值 (经过log1p标准化)
数据内容
- 特征:
genome_name: 基因组名称 (字符串)contig_name: 基因组片段名称 (字符串序列)protein_id: 蛋白质ID (字符串序列)protein_sequence: 蛋白质序列 (字符串序列)taxid: 分类ID (字符串)locus_tag: 基因座标签 (字符串序列)start: 起始位置 (整数序列)end: 终止位置 (整数序列)product: 产物描述 (字符串序列)
标签信息
- 二元标签文件:
binary_labels.csv(36种抗生素) - MIC回归标签文件:
mic_regression_labels.csv(56种抗生素)
数据集统计
- 训练集大小: 40,490,436,864字节
- 训练集样本数: 26,052
- 下载大小: 34,207,458,365字节
使用建议
-
加载方式: 推荐使用流式加载以避免内存错误 python from datasets import load_dataset ds = load_dataset("macwiatrak/bacbench-antibiotic-resistance-protein-sequences", split="train", streaming=True)
-
数据分割: 建议使用k-fold分割 (5等分)
- 二元标签: 使用
StratifiedKFold - 回归标签: 使用
KFold - 验证集占比: 20%
- 二元标签: 使用
相关资源
- DNA序列表示数据集
- GitHub代码库 (包含DNA和蛋白质语言模型嵌入代码)




