Skip to content
Get packs

PyTDC Therapeutics Data Toolkit

Discover, load, split and evaluate Therapeutics Data Commons datasets and benchmarks with PyTDC

Source Download skill
Repository
K-Dense-AI/scientific-agent-skills
Stars
55,780
Licence
MIT
Files
11 files

What it does

This skill teaches an AI agent how to work with the Therapeutics Data Commons (TDC) library, a collection of curated machine-learning datasets and benchmarks for drug discovery. The agent first looks up what tasks and datasets exist using metadata only, then plans a download and a task-appropriate train/test split, and only fetches data after the user approves the license, size and storage. It can also score predictions against benchmark metrics and audit a local data cache.

Great for

  • Researchers or students building machine-learning models on drug-discovery datasets
  • Data scientists who need reproducible splits and standard evaluation metrics for TDC benchmarks
  • Developers who want an agent to follow a disciplined discover-plan-download workflow

Skip it if

  • Users needing clinical or medical advice, since the datasets are research resources and not validated for patient decisions
  • Commercial projects without checking each dataset's license, because dataset terms vary from open Creative Commons to non-commercial or unspecified
  • Environments with limited disk space or no network access, because the install and dataset downloads can be hundreds of megabytes or more

Try asking

  • “List the ADME datasets available in PyTDC and show the official names”
  • “Plan a scaffold split for Caco2_Wang with seed 42 without downloading anything yet”
  • “Evaluate my prediction file against the admet_group benchmark metrics”
  • “Audit my .pytdc-data cache folder and report its size and file types”
Step-by-step guide: how to use PyTDC Therapeutics Data ToolkitInstall, first run, examples, tips and FAQs

Install in 3 steps

  1. 1

    Download the skill

    The ZIP contains the 11 original files, the MIT licence and an attribution note.

    pytdc.zip
  2. 2

    Add it to your agent

    Claude Code (all projects)

    mkdir -p ~/.claude/skills && unzip pytdc.zip -d ~/.claude/skills/

    Claude Code / Agent SDK (this project only)

    mkdir -p .claude/skills && unzip pytdc.zip -d .claude/skills/

    Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.

  3. 3

    Ask for the task

    No special command needed — the agent reads the skill's description and loads it when your request matches. Try one of the prompts above.

Get better results

  • Ask for a metadata-only discovery or split plan first, and review the license and size before approving any download
  • Use exact dataset and benchmark names returned by discovery rather than approximate names to avoid selecting the wrong data
  • Keep molecular oracle and remote docking use off unless you understand that molecular structures may be sent to external services

What's inside

  • references/datasets.md
  • references/oracles.md
  • references/sources.md
  • references/utilities.md
  • scripts/_common.py
  • scripts/benchmark_evaluation.py
  • scripts/cache_audit.py
  • scripts/discover_metadata.py
  • scripts/load_and_split_data.py
  • scripts/molecular_generation.py
  • SKILL.md

Redistributed unmodified from K-Dense-AI/scientific-agent-skills under the MIT licence. The assessment was drafted with AI from the skill's files on 2026-10-29. Skills can instruct your agent to run code — read SKILL.md and any scripts before installing.