OgBench: A Framework for Evaluating Graph Neural Networks on Omics Data

2026-05-15

AI Summary

Existing graph neural network (GNN) benchmarks predominantly focus on scenarios where the number of graphs far exceeds the number of nodes per graph, overlooking the typical “few samples, large graphs” (n ≪ p) setting prevalent in omics data. To address this gap, this work proposes OgBench, the first benchmarking framework tailored to this regime, which establishes a standardized end-to-end pipeline transforming raw omics data into diverse, feature-rich graphs. OgBench integrates classical GNNs, large-graph-optimized models, MLPs, and traditional machine learning methods. Empirical results reveal that state-of-the-art GNNs often fail to outperform simple baselines under this setting, challenging the common assumption that graph structure inherently improves predictive performance. OgBench provides a reliable benchmark and open-source ecosystem to foster the development of novel architectures for biological graph learning.

Abstract

Graph Neural Networks (GNNs) have become the dominant framework for inductive graph-level learning. Yet most benchmarks focus on the regime $n \gg p$, where the number of graphs $n$ greatly exceeds the number of nodes per graph $p$. This overlooks biological domains such as omics, which operate in the opposite $n \ll p$ regime, characterized by large graphs of genes, transcripts, or proteins across few patient samples. This raises the question: \textit{how do GNNs perform in this low-sample, high-node omics setting?} We introduce \texttt{OgBench} (Omics-Graph Bench), the first benchmarking platform for graph-level prediction in the $n \ll p$ regime characteristic of omics data. We provide a standardized, end-to-end modular infrastructure from raw omics data to families of featured graphs with varied structural properties. We benchmark classical GNNs, as well as GNNs designed for large graphs and omics applications, alongside MLPs and machine learning baselines to establish reference performances. Our results show that widely used GNNs often do not outperform simple MLPs and classical baselines. These findings challenge the prevailing assumption that graph structure inherently adds value in this domain, fostering a critical reassessment of current learning paradigms. Ultimately, by exposing these limitations, OgBench provides the open-source ecosystem necessary for the community to develop and validate novel architectures explicitly tailored for biological graphs. The code is available at https://github.com/geometric-intelligence/ogbench.

Problem

• Graph Neural Networks
• omics data
• low-sample high-node regime
• graph-level prediction
• benchmarking

Innovation

• Graph Neural Networks
• Omics Data
• Benchmarking
• Low-sample High-node Regime
• Biological Graphs