PharmaCPU3 credits

ADMET-AI v2 — Comprehensive Drug Safety Profiling

41 ADMET endpoints from Therapeutics Data Commons in one API call

ADMET-AI v2 uses Chemprop graph neural networks trained on 41 Therapeutics Data Commons (TDC) datasets to predict comprehensive absorption, distribution, metabolism, excretion, and toxicity properties. Unlike traditional rule-based ADMET tools, ADMET-AI learns complex structure-activity relationships directly from curated experimental data. Returns risk flags, drug-likeness assessment, and tissue-specific summaries.

$0.03

per API call

credits per call

/v1/pharma/admet-ai

API endpoint

Features

41 ADMET predictions in a single API call

Trained on Therapeutics Data Commons (TDC) benchmarks

Chemprop graph neural network + RDKit descriptors

Risk flag generation (hERG, AMES, DILI, CYP pan-inhibition)

Drug-likeness assessment (Lipinski, beyond Rule of 5)

Batch support up to 100 molecules per request

Quick Start

ADMET-AI v2 (Chemprop-RDKit) — Python Examplepython

import requests

API_KEY = "sk-sci-your-key-here"
url = "https://scirouter.ai/v1/pharma/admet-ai"

response = requests.post(url, json={
    "smiles": "CC(=O)Oc1ccccc1C(=O)O"  # Aspirin
}, headers={"Authorization": f"Bearer {API_KEY}"})

result = response.json()["data"]
print(f"Model: {result['model']}")
print(f"Endpoints: {result['num_endpoints']}")
print(f"hERG risk: {result['summary']['toxicity']['herg_risk']}")
print(f"CYP inhibitors: {result['summary']['metabolism']['cyp_inhibitors']}/5")
for flag in result['risk_flags']:
    print(f"  ⚠ {flag['property']}: {flag['detail']}")