An Analytical Neighborhood Enrichment Score for Spatial Omics

TL;DR

This paper introduces an analytical neighborhood enrichment score for spatial omics that approximates traditional permutation tests with high accuracy and significantly improves computational efficiency for large datasets.

Contribution

The study develops an analytical method for neighborhood enrichment testing in spatial omics, replacing computationally intensive permutation tests with a fast, accurate alternative.

Findings

High correlation (Pearson r ≥ 0.95) with traditional methods across datasets

Significant speed-up in analysis time for large-scale data

Effective application to large Xenium dataset

Abstract

The neighborhood enrichment test is used to quantify spatial enrichment and depletion between spatial points with categorical labels, which is a common data type in spatial omics. Traditionally, this test relies on a permutation-based Monte Carlo approach, which tends to be computationally expensive for large datasets. In this study, we present a modified version of the test that can be computed analytically. This analytical version showed a minimum Pearson correlation of 0.95 with the conventional Monte Carlo-based method across eight spatial omics datasets, but with substantial speed-ups. Additional experiments on a large Xenium dataset demonstrated the method's ability to efficiently analyze large-scale data, making it a valuable tool for analyzing spatial omics data.