README.md

metadata

license: mit

Dataset Card for ogbg-molpcba

Table of Contents

• Table of Contents
• Dataset Description
  • Dataset Summary
  • Supported Tasks and Leaderboards
• External Use
  • PyGeometric
• Dataset Structure
  • Data Properties
  • Data Fields
  • Data Splits
• Additional Information
  • Licensing Information
  • Citation Information
  • Contributions

Dataset Description

• Homepage
• Repository::
• Paper:: Open Graph Benchmark: Datasets for Machine Learning on Graphs (see citation)
• Leaderboard:: OGB leaderboard and Papers with code leaderboard

Dataset Summary

The ogbg-molpcba dataset is a small molecular property prediction dataset, adapted from MoleculeNet by teams at Stanford, to be a part of the Open Graph Benchmark.

Supported Tasks and Leaderboards

ogbg-molpcba should be used for molecular property prediction (with 128 properties to predict, not all present for all graphs), a binary classification task. The score used is Average Precision (AP) averaged over the tasks.

The associated leaderboards are here: OGB leaderboard and Papers with code leaderboard.

External Use

PyGeometric

To load in PyGeometric, do the following:

from datasets import load_dataset

from torch_geometric.data import Data
from torch_geometric.loader import DataLoader

dataset = load_dataset("graphs-datasets/ogbg-molpcba")
# For the train set (replace by valid or test as needed)
graphs_list_pygeometric = [Data(graph) for graph in dataset["train"]]
dataset_pygeometric = DataLoader(graphs_list_pygeometric)

Dataset Structure

Data Properties

property	value
scale	medium
#graphs	437,929
average #nodes	26.0
average #edges	28.1
average node degree	2.2
average cluster coefficient	0.002
MaxSCC ratio	0.999
graph diameter	13.6

Data Fields

Each row of a given file is a graph, with:

• x (list: #nodes x #node-features): nodes
• edge_index (list: 2 x #edges): pairs of nodes constituting edges
• edge_attr (list: #edges x #edge-features): for the aforementioned edges, contains their features
• y (list: 1 x #labels): contains the number of labels available to predict (here 128 labels, equal to zero, one, or Nan if the property is not relevant for the graph)
• num_nodes (int): number of nodes of the graph

Data Splits

This data comes from the PyGeometric version of the dataset provided by OGB, and follows the provided data splits. This information can be found back using

from ogb.graphproppred import PygGraphPropPredDataset

dataset = PygGraphPropPredDataset(name = 'ogbg-molhiv')

split_idx = dataset.get_idx_split() 
train = dataset[split_idx['train']] # valid, test

Additional Information

Licensing Information

The dataset has been released under MIT license.

Citation Information

@inproceedings{hu-etal-2020-open,
  author    = {Weihua Hu and
               Matthias Fey and
               Marinka Zitnik and
               Yuxiao Dong and
               Hongyu Ren and
               Bowen Liu and
               Michele Catasta and
               Jure Leskovec},
  editor    = {Hugo Larochelle and
               Marc Aurelio Ranzato and
               Raia Hadsell and
               Maria{-}Florina Balcan and
               Hsuan{-}Tien Lin},
  title     = {Open Graph Benchmark: Datasets for Machine Learning on Graphs},
  booktitle = {Advances in Neural Information Processing Systems 33: Annual Conference
               on Neural Information Processing Systems 2020, NeurIPS 2020, December
               6-12, 2020, virtual},
  year      = {2020},
  url       = {https://proceedings.neurips.cc/paper/2020/hash/fb60d411a5c5b72b2e7d3527cfc84fd0-Abstract.html},
}

Contributions

Thanks to @clefourrier for adding this dataset.