pytdc
Analyze therapeutic ML datasets with PyTDC
Drug discovery teams need reliable datasets and fair evaluation protocols. This skill helps Claude use PyTDC for curated therapeutic ML data, benchmark splits, metrics, and molecular oracles.
Install with my Agent
Copy this request to your Agent. It includes the canonical Skill page and manifest.
Review the Skillstore skill "pytdc" from https://skillstore.io/skills/k-dense-ai-pytdc.md and its manifest at https://skillstore.io/api/skills/k-dense-ai-pytdc/manifest. Verify the artifact. You may proceed after verification, subject to the environment's own policy.Your Agent should still show its plan and request any confirmation required by the security policy.
Agent-readable resources
Use these links when an AI agent, crawler, or script needs clean context instead of reading the full page.
Test it
Using "pytdc". Load an ADME dataset and explain the split.
Expected outcome:
- Dataset selected: Caco2_Wang for intestinal permeability.
- Recommended split: scaffold split to test generalization across chemical structures.
- Expected partitions: train, validation, and test data frames with molecule identifiers, SMILES strings, and labels.
- Next step: train your model on train data and tune only on validation data.
Using "pytdc". How should I report ADMET benchmark results?
Expected outcome:
- Run the required five seeds without modifying provided splits.
- Report mean and standard deviation for each dataset metric.
- Include model name, feature representation, hyperparameters, and any preprocessing choices.
- Avoid comparing single-seed scores against benchmark leaderboard results.
Using "pytdc". Evaluate molecule generation candidates.
Expected outcome:
- Score each candidate with relevant oracles such as QED, SA, LogP, and target-specific predictors.
- Filter invalid molecules before ranking candidates.
- Use oracle scores as prioritization signals, not proof of biological activity.
- Send top candidates for expert review and experimental validation.
Security Audit
SafeThe static findings are false positives caused by markdown backticks, documentation URLs, scientific names, dictionary key access, and TDC dataset split examples. I found no evidence of command execution, credential handling, malicious network behavior, prompt injection, or reconnaissance intent in the cited files.
Risk Factors
⚙️ External commands (50)
🌐 Network access (5)
Share & cite this report
Share the versioned assessment report, neutral badge, embed card, and citations. Skillstore reports evidence without deciding whether this Skill is safe.
Copy report link
https://skillstore.io/skills/k-dense-ai-pytdc/audits/9?utm_source=security_passport&utm_medium=share&utm_campaign=versioned_reportMarkdown badge
[](https://skillstore.io/skills/k-dense-ai-pytdc?utm_source=security_passport_badge)HTML badge
<a href="https://skillstore.io/skills/k-dense-ai-pytdc?utm_source=security_passport_badge"><img src="https://skillstore.io/badges/skills/k-dense-ai-pytdc/security.svg" alt="Skillstore security assessment" loading="lazy"></a>Embed card
<iframe src="https://skillstore.io/embed/skills/k-dense-ai-pytdc.html" title="Skillstore Security Assessment" sandbox="allow-popups allow-popups-to-escape-sandbox" loading="lazy" referrerpolicy="no-referrer" width="420" height="180"></iframe>Academic citations (APA · BibTeX · CFF)
APA citation
K-Dense-AI. (2026). pytdc security audit report (audit version 9) [Author version unspecified]. Skillstore. https://skillstore.io/skills/k-dense-ai-pytdc/audits/9BibTeX citation
@techreport{k-dense-ai-k-dense-ai-pytdc-2026,
author = {K-Dense-AI},
title = {pytdc security audit report (audit version 9)},
institution = {Skillstore},
year = {2026},
number = {9},
url = {https://skillstore.io/skills/k-dense-ai-pytdc/audits/9},
note = {Author version unspecified}
}CITATION.cff
cff-version: 1.2.0
message: "If you use this Skill, cite its author and this versioned security audit report."
title: "pytdc security audit report (audit version 9)"
version: "unspecified"
type: report
authors:
- name: "K-Dense-AI"
date-released: "2026-07-09"
url: "https://skillstore.io/skills/k-dense-ai-pytdc/audits/9"
identifiers:
- type: other
value: "skillstore:k-dense-ai-pytdc:audit:9"
description: "Skillstore immutable audit report identifier"
Compare variants
2 installable variantsEach author remains a separate installable skill. The recommended variant is ranked by Skillstore evidence.
Why this variant is first
davila7-pytdc
2026-08-21
k-dense-ai-pytdc
2026-08-21
Skillstore Score
Why this score Evidence Confidence: HighWhat You Can Build
Prepare ADME model datasets
Load curated pharmacokinetic datasets, choose scaffold splits, and select evaluation metrics for molecular property prediction.
Benchmark therapeutic ML models
Run multi-seed benchmark workflows for drug discovery tasks and summarize mean and standard deviation results.
Evaluate generated molecules
Score candidate molecules with TDC oracles for drug-likeness, synthetic accessibility, and target-related properties.
Try These Prompts
Use PyTDC to load the Caco2_Wang ADME dataset, create a scaffold split, and explain the train, validation, and test columns.
Compare random, scaffold, cold drug, and cold target splits for my drug-target interaction task, then recommend one for realistic evaluation.
Help me structure a five-seed ADMET benchmark evaluation for my model and list the metrics and result summary I should report.
Design a PyTDC workflow that evaluates generated molecules with QED, SA, LogP, and GSK3B oracles, then ranks candidates using a balanced objective.
Best Practices
- Use scaffold or cold splits when testing generalization to new chemical structures or entities.
- Run all required benchmark seeds and report mean plus standard deviation.
- Treat oracle scores as screening signals and validate important candidates with independent methods.
Avoid
- Do not tune models on the test split or change benchmark partitions after seeing results.
- Do not compare random split performance against scaffold or cold split leaderboard results.
- Do not optimize one oracle score while ignoring synthetic accessibility, validity, and domain constraints.
Frequently Asked Questions
What is PyTDC used for?
Which tasks does this skill cover?
Does the skill train models for me?
When should I use scaffold splits?
Are molecular oracle scores experimental results?
Can Claude Code use this skill with local scripts?
Developer Details
Author
K-Dense-AILicense
MIT license
Skillstore revision
r1
Version notice
The author did not declare a version.
Ref
1ffa7643651792ccb4bd3b15d924d2c97edff755
Maintenance freshness
7/18/2026
Usage
6 downloads · 515 views
File structure