ASAP-Polaris-OpenADMET Antiviral Challenge Dataset: sample notebook
A worked notebook for ASAP-Polaris-OpenADMET Antiviral Challenge Dataset, with 4 code cells and the output each one printed. On Scigantic it opens in a hosted JupyterLab with the dataset already mounted, so nothing is downloaded first.
- Provider
- OpenADMET
- License
- mit
- Format
- CSV (in sampled files)
- Stored at
- Hugging Face (https://huggingface.co/datasets/openadmet/ASAP_Polaris_OpenADMET_challenge)
- Access
- Free and public. No account with the data provider is needed.
The notebook
ASAP-Polaris-OpenADMET Antiviral Challenge Dataset — Getting Started
Open antiviral structure-activity data from ASAP Discovery, spanning potency, ADMET, and X-ray crystal structures for the same compounds against two main-protease targets: SARS-CoV-2 and MERS-CoV Mpro. That combination — activity, ADMET, and structure on one compound set — is rare in a public dataset.
The data ships as flat CSVs at the repo root rather than one HF table, so this notebook reads them directly.
import pandas as pd
admet = pd.read_csv("hf://datasets/openadmet/ASAP_Polaris_OpenADMET_challenge/ADMET.csv")
pot = pd.read_csv("hf://datasets/openadmet/ASAP_Polaris_OpenADMET_challenge/Potency.csv")
print(f"ADMET: {admet.shape}, Potency: {pot.shape}")
pot.head()
Output:
ADMET: (560, 8), Potency: (1328, 5)
CXSMILES Molecule Name Set \
0 COC[C@]1(C)C(=O)N(C2=CN=CC3=CC=CC=C23)C(=O)N1C... ASAP-0000141 Train
1 C=C(CN1CCC2=C(C=C(Cl)C=C2)C1C(=O)NC1=CN=CC2=CC... ASAP-0000142 Train
2 CNC(=O)CN1C[C@]2(C[C@H](C)N(C3=CN=CC=C3C3CC3)C... ASAP-0000143 Train
3 C=C(CN1CCC2=C(C=C(Cl)C=C2)C1C(=O)NC1=CN=CC2=CC... ASAP-0000144 Train
4 C=C(CN1CCC2=C(C=C(Cl)C=C2)C1C(=O)NC1=CN=CC2=CC... ASAP-0000145 Train
pIC50 (MERS-CoV Mpro) pIC50 (SARS-CoV-2 Mpro)
0 4.19 NaN
1 4.92 5.29
2 4.73 NaN
3 4.90 6.11
4 4.81 5.62
Two targets, not one
Potency is split into per-target columns and is sparsely populated — many compounds have only one target measured. Compare the two distributions rather than assuming one "potency" column.
import matplotlib.pyplot as plt
fig, ax = plt.subplots(figsize=(7, 4))
pot["pIC50 (SARS-CoV-2 Mpro)"].dropna().hist(bins=25, alpha=0.6, label="SARS-CoV-2 Mpro", ax=ax, color="steelblue")
pot["pIC50 (MERS-CoV Mpro)"].dropna().hist(bins=25, alpha=0.6, label="MERS-CoV Mpro", ax=ax, color="darkorange")
ax.set_xlabel("pIC50"); ax.legend(); ax.set_title("Potency by target")
plt.tight_layout()
Potency and ADMET aren't pre-joined
Merge on Molecule Name yourself.
merged = pot.merge(admet, on="Molecule Name", how="left", suffixes=("_pot", "_admet"))
merged[["Molecule Name", "pIC50 (SARS-CoV-2 Mpro)", "LogD", "KSOL"]].head()
Output:
Molecule Name pIC50 (SARS-CoV-2 Mpro) LogD KSOL 0 ASAP-0000141 NaN NaN NaN 1 ASAP-0000142 5.29 NaN NaN 2 ASAP-0000143 NaN NaN NaN 3 ASAP-0000144 6.11 NaN NaN 4 ASAP-0000145 5.62 NaN NaN
What the most potent SARS-CoV-2 inhibitors look like
Structures here are CXSMILES (ChemAxon extended SMILES, with stereo annotations like |a:16|) — RDKit's MolFromSmiles handles the CX block directly.
from rdkit import Chem
from rdkit.Chem import Draw
def draw(smiles_list, legends, per_row=4):
mols, labels = [], []
for s, l in zip(smiles_list, legends):
m = Chem.MolFromSmiles(s)
if m is not None:
mols.append(m)
labels.append(l)
return Draw.MolsToGridImage(mols, molsPerRow=per_row, subImgSize=(220, 180), legends=labels)
top = pot.dropna(subset=["pIC50 (SARS-CoV-2 Mpro)"]).nlargest(8, "pIC50 (SARS-CoV-2 Mpro)")
draw(top["CXSMILES"], [f"pIC50={v:.1f}" for v in top["pIC50 (SARS-CoV-2 Mpro)"]])
Where to go next
- structures.tar.gz at the repo root holds the paired X-ray crystal structures (Train/Test filename prefix).
- HLM/MLM columns exist in ADMET.csv but are essentially empty in this release.
- Train/test membership lives in the Set column.
- Open this archive's schema card for the full column reference.
Run it
Open this notebook to run it, see the dataset page for the full description, or browse the full catalog.