OpenADMET Octant CYP Inhibition & Reactivity Dataset: sample notebook

A worked notebook for OpenADMET Octant CYP Inhibition & Reactivity 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
apache-2.0
Format
TSV (in sampled files)
Stored at
Hugging Face (https://huggingface.co/datasets/openadmet/Octant_CYP_inhibition_reactivity_blog_release)
Access
Free and public. No account with the data provider is needed.

The notebook

OpenADMET Octant CYP Inhibition & Reactivity — Getting Started

Self-consistent, multi-endpoint cytochrome P450 (CYP3A4) data generated by Octant Bio on a single platform under controlled conditions. Public CYP datasets (PubChem bioassays, Veith 2009) are pooled from many labs and effectively binary; this one ships full dose-response curves plus real QC flags for roughly 1,200 compounds — much cleaner, if you respect the QC columns.

This notebook loads the inhibition track, shows why the QC filter matters, and looks at the separate reactivity assay.

from datasets import load_dataset
import pandas as pd

df = load_dataset("openadmet/Octant_CYP_inhibition_reactivity_blog_release", "inhibition", split="train").to_pandas()
print(f"{len(df)} compounds")
df.head()

Output:

1340 compounds
            ocnt_batch  CYP3A4_pIC50  CYP3A4_pIC50_se  CYP3A4_pIC50_ci_lower  \
0  OCNT-0497356-AA-001      7.467446         0.019508               7.430739   
1  OCNT-0497490-AA-001      7.311531         0.016860               7.277453   
2  OCNT-0475246-AA-001      7.264564         0.043869               7.173135   
3  OCNT-0456376-AA-001      7.153188         0.018771               7.114670   
4  OCNT-0456186-AA-001      7.149408         0.030117               7.090624   

   CYP3A4_pIC50_ci_upper  slope_log2  emax_log2fc activity_status  \
0               7.505657   -0.298057    -5.221625             YES   
1               7.345002    0.394112    -4.103657             YES   
2               7.345680   -0.474457    -5.379198             YES   
3               7.189992   -0.482009    -5.139855             YES   
4               7.208831    0.337081    -4.110778             YES   

...

Filter on QC before trusting a pIC50

drc_qc_status marks whether the dose-response curve fit cleanly. Rows that fail QC carry an unreliable pIC50 — worth seeing side by side before deciding whether to drop them.

print(df["drc_qc_status"].value_counts())

import matplotlib.pyplot as plt
fig, ax = plt.subplots(figsize=(7, 4))
for status, color in [("PASS", "steelblue"), ("FAIL", "indianred")]:
    sub = df.loc[df["drc_qc_status"] == status, "CYP3A4_pIC50"].dropna()
    ax.hist(sub, bins=25, alpha=0.6, label=f"{status} (n={len(sub)})", color=color)
ax.set_xlabel("CYP3A4 pIC50"); ax.legend(); ax.set_title("Inhibition potency by QC status")
plt.tight_layout()

Output:

drc_qc_status
PASS    1139
FAIL     201
Name: count, dtype: int64

Reactivity is a different assay — don't mix it with inhibition

reactivity measures substrate depletion / reaction phenotyping for CYP3A4 and CYP2J2 (pct_remaining, log2fc), not inhibition potency. Same compounds, different biology.

reactivity = load_dataset("openadmet/Octant_CYP_inhibition_reactivity_blog_release", "reactivity", split="train").to_pandas()
reactivity[["ocnt_batch", "enzyme", "pct_remaining", "log2fc"]].head()

Output:

            ocnt_batch  enzyme  pct_remaining    log2fc
0  OCNT-0017486-AA-002  CYP2J2     111.440990  0.156272
1  OCNT-0017486-AA-002  CYP3A4     103.977294  0.056266
2  OCNT-0021863-AA-002  CYP2J2      25.318560 -1.981502
3  OCNT-0021863-AA-002  CYP3A4      34.282450 -1.544324
4  OCNT-0022119-AA-003  CYP2J2     103.736770  0.052927

What the strongest QC-passing inhibitors look like

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)

clean = df[df["drc_qc_status"] == "PASS"].dropna(subset=["CYP3A4_pIC50"])
top = clean.nlargest(8, "CYP3A4_pIC50")
draw(top["standardized_smiles"], [f"pIC50={v:.1f}" for v in top["CYP3A4_pIC50"]])

Where to go next

Run it

Open this notebook to run it, see the dataset page for the full description, or browse the full catalog.