Relevant Datasets and Software Used for Paper "KGML-xDTD: A Knowledge Graph-based Machine Learning Framework for Drug Treatment Prediction and Mechanism Description"

This repository contains relevant datasets and software used in a paper "KGML-xDTD: A Knowledge Graph-based Machine Learning Framework for Drug Treatment Prediction and Mechanism Description". They are used to run the code of <em>KGML-xDTD </em>stored on Github and support the results of this paper. <strong>About the datasets</strong> 1. <em>bkg_rtxkg2c_v2.7.3.tar.gz</em> This tar.gz file contains three sub-folders: tsv_files, scripts, and relevant_dbs. The "tsv_files" sub-folder has the input files that the neo4j software uses. The "scripts" sub-folder contains a shell script with a relevant python script to construct the biomedical knowledge graph. The "relevant_dbs" sub-folder stores two auxiliary databases that <em>KGML-xDTD</em> needs to use. 2. <em>indication_paths.yaml</em> This file contains the DrugMechDB MOA paths that we used to evaluate the predicted MOA paths by <em>KGML-xDTD. </em>It is downloaded from the official GitHub repository of DrugMechDB. 3. <em>training_data.tar.gz</em> This tar.gz file contains the processed training data of four data sources (e.g., MyChem, SemMedDB, NDF-RT, RepoDB) mentioned in the paper. These processed drug-disease pairs have been matched to the identifiers of biological entities used in our biomedical knowledge graph and respectively split into true positive (tp) sets and true negative (tn) sets. We also provide the names of these drug identifiers and disease identifiers under a sub-folder "translated _to_name". <strong>About the software</strong> <em>neo4j-community-3.5.26.tar.gz</em> This tar.gz is the Neo4j community version 3.5.26 downloaded from Neo4j Download Center. Although the newer versions are available, due to their big changes in the Neo4j setting that are not compatible with our scripts on Github, we provide the version that we used in our research. If you would like to use the newer version, modifications to our script will be required to import the biomedical knowledge graph into your local Neo4j database with the new setting.

本仓库包含论文《KGML-xDTD：一款基于知识图谱（Knowledge Graph）的药物治疗预测与机制阐释机器学习（Machine Learning）框架》中所使用的相关数据集与软件资源，用于运行GitHub上存储的*KGML-xDTD*代码，并支撑该论文的相关结果。

**关于数据集**
1. *bkg_rtxkg2c_v2.7.3.tar.gz*
该tar.gz压缩包内含三个子文件夹：`tsv_files`、`scripts`与`relevant_dbs`。其中`tsv_files`子文件夹存储Neo4j软件所需的输入文件；`scripts`子文件夹包含用于构建生物医学知识图谱（biomedical knowledge graph）的Shell脚本与配套Python脚本；`relevant_dbs`子文件夹则存放*KGML-xDTD*运行所需的两个辅助数据库。
2. *indication_paths.yaml*
该文件包含我们用于评估*KGML-xDTD*所预测的作用机制（Mechanism of Action, MOA）路径的DrugMechDB MOA路径集，其源自DrugMechDB的官方GitHub仓库。
3. *training_data.tar.gz*
该tar.gz压缩包内含论文中提及的四个数据源（如MyChem、SemMedDB、NDF-RT、RepoDB）的预处理训练数据。这些经过处理的药物-疾病对已与我们生物医学知识图谱中使用的生物实体标识符完成匹配，并分别划分为真阳性（true positive, tp）集合与真阴性（true negative, tn）集合。我们还在`translated_to_name`子文件夹中提供了这些药物标识符与疾病标识符的对应名称。

**关于软件**
*neo4j-community-3.5.26.tar.gz*
该tar.gz压缩包为从Neo4j下载中心获取的Neo4j社区版3.5.26。尽管已有更新版本推出，但由于新版本在Neo4j配置上存在较大改动，与我们GitHub上的脚本不兼容，因此我们提供了研究中实际使用的版本。若您希望使用更新版本，则需对我们的脚本进行修改，才能将生物医学知识图谱导入至您本地的新版Neo4j数据库中。