Dataset Viewer
Auto-converted to Parquet Duplicate
The dataset viewer is not available for this split.
The number of columns (1652) exceeds the maximum supported number of columns (1000). This is a current limitation of the datasets viewer. You can reduce the number of columns if you want the viewer to work.
Error code:   TooManyColumnsError

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

Bober KRFP PubChem Patterns

Molecule-ID datasets for 1,652 acyclic, relaxed KRFP SMARTS patterns, built from 105,842,232 PubChem records. Each pattern has up to 5 million IDs per class.

Files

File Contents
KRFPxxxx.npz Three uint32 arrays: positive, tversky, negative
patterns.json Relaxed SMARTS queries used to label matches
pre_patterns.json Original SMARTS queries used to construct similarity prototypes
smiles.txt One SMILES per global row ID
smiles_offsets.npy Byte offsets into smiles.txt (uint64, length N + 1)

Labels: positive matches the relaxed SMARTS; tversky does not match but passes a Morgan-fingerprint Tversky similarity threshold (radius 2, 2048 bits, α=1, β=0, score ≥0.5); negative is a sampled nonmatch. Selected classes for each pattern are disjoint.

IDs are zero-based positions in the shared SMILES file, not PubChem CIDs.

Use

import numpy as np
from huggingface_hub import hf_hub_download

repo = "WiktorMatuszek/bober_patterns"

def get_file(name):
    return hf_hub_download(repo, name, repo_type="dataset")

with np.load(get_file("KRFP0001.npz")) as data:
    positive = data["positive"]
    tversky = data["tversky"]
    negative = data["negative"]

offsets = np.load(get_file("smiles_offsets.npy"), mmap_mode="r")
with open(get_file("smiles.txt"), "rb") as f:
    idx = int(positive[0])
    f.seek(int(offsets[idx]))
    smiles = f.read(int(offsets[idx + 1] - offsets[idx])).decode().strip()

print(idx, smiles)

Download only the required pattern files; the shared smiles.txt and offset index are needed only to resolve IDs.

Notes

SMARTS are automatically relaxed; Some patterns have few or no positive examples. Duplicates are removed by exact SMILES text, not canonical structure.

Downloads last month
19