Therapeutics Data Commons

实用信息

内容范围

• 文中说明，TDC 包含 66 个可供 AI 使用的数据集，分布在 22 个学习任务上，覆盖安全有效药物的发现与开发。
  arxiv.org · 2026-11-13

  To date, TDC includes 66 AI-ready datasets spread across 22 learning tasks and spanning the discovery and development of safe and effective medicines.

  https://arxiv.org/abs/2102.09548 · preprint · 核实于 2026-11-13

• 文中说明，TDC 覆盖靶点发现、活性筛选、有效性、安全性和生产等学习任务，涉及小分子、抗体和疫苗等生物医学产品。
  tdcommons.ai · 2026-11-13

  TDC covers a wide range of learning tasks, including target discovery, activity screening, efficacy, safety, and manufacturing across biomedical products, including small molecules, antibodies, and vaccines.

  https://tdcommons.ai/ · 数据集页面 · 核实于 2026-11-13

标识符

• dataset name within a task class文中显示，通过名称 HIA_Hou 从单实例预测问题下的 ADME 治疗学习任务中获取一个数据集。
  raw.githubusercontent.com · 2026-11-13

  For a concrete example, to obtain the HIA dataset from the ADME therapeutic learning task in the single-instance prediction problem:

  https://raw.githubusercontent.com/mims-harvard/TDC/HEAD/README.md · README · 核实于 2026-11-13

取用方式

• 仅网页界面github.com

  条款同本卡许可。

版本

• 最新TDC 以 Python 包 PyTDC 的版本号区分；PyPI 最新版本为 1.1.15，上传于 2025-03-31。未见独立的数据版本编号。
  pypi.org · 2026-11-13

  "upload_time":"2025-03-31T15:20:14"

  https://pypi.org/pypi/PyTDC/json · official-release-notes · 核实于 2026-11-13

• 更新周期文中说明，TDC 会定期更新，加入新的数据集和任务，例如抗体疗法和基因编辑。
  tdcommons.ai · 2026-11-13

  TDC is regularly updated with novel datasets and tasks, such as antibody therapeutics and gene editing.

  https://tdcommons.ai/overview/ · 数据集页面 · 核实于 2026-11-13

已知缺陷与常见误用

• 作者在文中写道：按他列出的标准，TDC 与 MoleculeNet“同样有缺陷”。
  practicalcheminformatics.blogspot.com · 2026-11-13

  other widely used benchmarks like the Therapeutic Data Commons (TDC) are equally flawed.

  https://practicalcheminformatics.blogspot.com/2023/08/we-need-better-benchmarks-for-machine.html · practitioner-article · 核实于 2026-11-13

• Walters：MoleculeNet 的 BBB 数据集含 59 个重复结构，其中 10 对标签相互矛盾，另有 11 个 RDKit 无法解析的 SMILES。
  practicalcheminformatics.blogspot.com · 2026-11-13

  First, the dataset contains 59 duplicate structures. This is not the sort of thing we want to have in a benchmark. But wait, it gets worse, 10 of these duplicate structures have different labels.

  https://practicalcheminformatics.blogspot.com/2023/08/we-need-better-benchmarks-for-machine.html · practitioner-article · 核实于 2026-11-13

• Walters：MoleculeNet 的 BACE 数据集中 71% 的分子至少有一个未定义的立体中心。
  practicalcheminformatics.blogspot.com · 2026-11-13

  71% of the molecules have at least one undefined stereocenter, 222 molecules have 3 undefined stereocenters, and one molecule has a whopping 12 undefined stereocenters.

  https://practicalcheminformatics.blogspot.com/2023/08/we-need-better-benchmarks-for-machine.html · practitioner-article · 核实于 2026-11-13

适合的任务

• 文中说明，TDC 是一个用于访问和评估 AI 方法的资源，支持面向药物发现应用的 AI 方法开发。
  tdcommons.ai · 2026-11-13

  Therapeutics Data Commons is a resource to access and evaluate AI methods, supporting the development of AI methods, with a strong bent towards establishing the foundation of which AI methods are most suitable for drug discovery applications and why.

  https://tdcommons.ai/ · 数据集页面 · 核实于 2026-11-13

• TDC 作者称其实验显示最强算法在真实分布偏移与对新数据的泛化上仍有不足。
  arxiv.org · 2026-11-13

  demonstrating that even the strongest algorithms fall short of solving key therapeutics challenges, including real dataset distributional shifts, multi-scale modeling of heterogeneous data, and robust generalization to novel data points.

  https://arxiv.org/abs/2102.09548 · preprint · 核实于 2026-11-13

• Walters：MoleculeNet 的生理类数据集（BBB、毒性）终点复杂，不太适合做基准比较。
  practicalcheminformatics.blogspot.com · 2026-11-13

  The complexity of these endpoints makes the data less than ideal for benchmark comparisons.

  https://practicalcheminformatics.blogspot.com/2023/08/we-need-better-benchmarks-for-machine.html · practitioner-article · 核实于 2026-11-13