PharMolixPharMolix/OpenBioMed/skills/admet-prediction/SKILL.md

admet-prediction

Skill1.1k starsChanged 8 months agoMITView on GitHub

Predict comprehensive ADMET (Absorption, Distribution, Metabolism, Excretion, Toxicity) properties for drug candidate molecules using GraphMVP ensemble models. Use this skill when: (1) Predicting blood-brain barrier penetration, (2) Assessing side effect profiles, (3) Estimating Caco-2 permeability, half-life, or LD50 toxicity, (4) Evaluating drug-likeness and safety of molecules.

4 other projects use a version of this.

Other versions4 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 4
K-Dense-AI/drug-discovery-agent-skills★ 54same name
pradyumnasagar/open-research-skills★ 8same name
BuildHarbor/sciagent-skills★ 1same name
Pratikrishi97/sciagent-skills★ 1same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository45 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- antibody-design-iggmSkill
- antibody-structure-prediction-tfoldSkill
- binding-affinity-prediction-prodigySkill
- biomedical-literature-searchSkill
- biomed-skill-creatorSkill
- biomed-skill-routerSkill
- cellxgene-census-querySkill
- chembl-querySkill
- disease-drug-intelligenceSkill
- drug-candidate-discoverySkill
- drug-drug-interaction-analysisSkill
- drug-lead-analysisSkill
- functional-protein-designSkill
- iupac-name-identification-biot5Skill
- kegg-querySkill
- molecule-biochemical-significance-query-biot5Skill
- mutation-design-aavSkill
- mutation-design-gfpSkill
- ppi-string-querySkill
- protein-function-predictionSkill
- protein-ligand-binding-analysis-plipSkill
- protein-mutation-analysisSkill
- protein-structure-design-boltzgenSkill
- protein-subcellular-localization-prediction-biot5Skill
- pubchem-querySkill
- retrosynthesis-planningSkill
- similar-protein-retrievalSkill
- single-cell-atac-seq-peak-calling-annotaionSkill
- single-cell-atac-seq-qc-processingSkill
- single-cell-foundation-model-geneformerSkill
- single-cell-foundation-model-langcellSkill
- single-cell-foundation-model-scgptSkill
- single-cell-multi-omics-analysis-scviSkill
- single-cell-multi-omics-data-harmonizationSkill
- single-cell-proteomics-data-processingSkill
- single-cell-proteomics-peptide-identificationSkill
- single-cell-scrna-seq-analysis-scanpySkill
- single-cell-foundation-model-stofmSkill
- spatial-transcriptomics-spatial-data-ioSkill
- structure-prediction-boltz-2Skill
- target-based-lead-designSkill
- target-drug-reportSkill
- text-based-molecule-editingSkill
- uniprot-querySkill

The file

We split this skill into the rules it gives an agent

3 rules. 0 are common, 2 are distinctive, and 1 is very specific to this project.

0
Common
Most of the 4 edited versions we read keep these.
2
Distinctive
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.

Common
Distinctive
Very specific

Distinctive:
Create molecule from SMILES string.
line 32
Very specific:
Solution: Ensure checkpoints exist in ./checkpoints/server/:
line 123
Distinctive:
Solution: Use CPU instead:
line 132

“Common” is measured: kept by at least 6 in 10 of the 4 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.

Files

SKILL.md

ADMET Prediction

Predict comprehensive ADMET properties for drug candidate molecules using GraphMVP ensemble models.

When to Use

User asks to predict ADMET properties for a molecule
User provides a drug candidate and wants safety assessment
User needs blood-brain barrier penetration prediction
User wants to evaluate toxicity (LD50) or side effects (SIDER)
User requests pharmacokinetic properties (half-life, Caco-2)

Workflow

Step 1: Load Molecule

Create molecule from SMILES string.

from open_biomed.data import Molecule
molecule = Molecule.from_smiles("CC(=O)OC1=CC=CC=C1C(=O)O")  # Aspirin

Step 2: Build ADMET Pipeline

Initialize ensemble pipeline with all GraphMVP checkpoints.

from open_biomed.core.pipeline import InferencePipeline, EnsemblePipeline

pipelines = {
    "BBBP": InferencePipeline(
        task="molecule_property_prediction", model="graphmvp",
        model_ckpt="./checkpoints/server/graphmvp-BBBP.ckpt",
        additional_config="./configs/dataset/bbbp.yaml", device="cuda:0"),
    "SIDER": InferencePipeline(
        task="molecule_property_prediction", model="graphmvp",
        model_ckpt="./checkpoints/server/graphmvp-SIDER.ckpt",
        additional_config="./configs/dataset/sider.yaml", device="cuda:0"),
    # See examples/basic_example.py for full pipeline setup
}
pipeline = EnsemblePipeline(pipelines)

Step 3: Run Predictions

Execute predictions for each ADMET property.

# BBB penetration
bbb_result = pipeline.run(molecule=molecule, task="BBBP")

# Side effects (27 categories)
sider_result = pipeline.run(molecule=molecule, task="SIDER")

# Regression properties
caco2_result = pipeline.run(molecule=molecule, task="caco2_wang")
half_life_result = pipeline.run(molecule=molecule, task="half_life_obach")
ld50_result = pipeline.run(molecule=molecule, task="ld50_zhu")

Expected Outputs

Task	Output Type	Description
BBBP	float [0-1]	Probability of BBB penetration
SIDER	list[27 floats]	Side effect probabilities per category
caco2_wang	float	Log permeability (cm/s)
half_life_obach	float	Log half-life (hours)
ld50_zhu	float	Log LD50 (mg/kg)

Interpretation Guide

BBB Penetration

Value	Interpretation
> 0.5	Likely crosses BBB
< 0.5	Unlikely to cross BBB

Caco-2 Permeability

Value (log cm/s)	Interpretation
> -5	High absorption
-6 to -5	Moderate absorption
< -6	Low absorption

LD50 Toxicity

Value (log mg/kg)	Toxicity Level
< 1	Highly toxic (<10 mg/kg)
1-2	Moderately toxic (10-100 mg/kg)
2-3	Slightly toxic (100-1000 mg/kg)
> 3	Low toxicity (>1000 mg/kg)

SIDER Side Effects

Values range 0-1. Categories with > 0.7 indicate high risk of that side effect.

Error Handling

Checkpoint Not Found

Symptom: FileNotFoundError: graphmvp-*.ckpt

Solution: Ensure checkpoints exist in ./checkpoints/server/:

ls checkpoints/server/graphmvp-*.ckpt

CUDA Out of Memory

Symptom: RuntimeError: CUDA out of memory

Solution: Use CPU instead:

# Change device from "cuda:0" to "cpu"
device="cpu"

Invalid SMILES

Symptom: Molecule fails to parse

Solution: Validate SMILES format or use molecule name lookup via PubChem.

Example

Input: aspirin (CC(=O)OC1=CC=CC=C1C(=O)O)

Output:
  BBB Penetration: 0.19 (does NOT cross BBB)
  Caco-2: -4.68 (moderate absorption)
  Half-life: -7.06 (short half-life)
  LD50: 2.06 (moderate toxicity ~115 mg/kg)

  Top Side Effects:
    Skin disorders: 0.80
    Nervous system: 0.78
    Gastrointestinal: 0.78

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

It worked
Partly
It didn't

What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

Check your setup, free