A sulfatide-centered ultra-high-resolution magnetic resonance MALDI imaging benchmark dataset for MS1-based lipid annotation tools

TL;DR

This paper introduces a high-resolution benchmark dataset for lipid annotation tools using mouse model data of a human disease, enabling better validation of computational methods.

Contribution

The paper provides two well-characterized, biology-driven MR-MSI datasets for benchmarking lipid annotation tools with ultra-high-resolution data.

Findings

The dataset includes ultra-high-resolution MR-MSI data with isotopic fine structure analysis for improved annotation confidence.

Manual sulfatide annotations were compared with decoy database-controlled annotations in Metaspace, showing 67 annotations at a false discovery rate <10%.

The datasets can be used to validate spatial biomarker discovery pipelines and study sulfatide metabolism.

Abstract

Spatial omics techniques are indispensable for studying complex biological systems and for the discovery of spatial biomarkers. While several current matrix-assisted laser desorption/ionization mass spectrometry imaging (MSI) instruments are capable of localizing numerous metabolites at high spatial and spectral resolution, most MSI data are acquired at the MS1 level only. Assigning molecular identities based on MS1 data presents significant analytical and computational challenges, as the inherent limitations of MS1 data preclude confident annotations beyond the sum formula level.

To enable future advancements of computational lipid annotation tools, well-characterized benchmark—or ground-truth—datasets are crucial, which exceed the scope of synthetic data or data derived from mimetic tissue models. To this end, we provide 2 sulfatide-centered, biology-driven magnetic resonance MSI (MR-MSI) datasets at different mass resolving powers that characterize lipids in a mouse model of human metachromatic dystrophy. These data include an ultra-high-resolution (R ∼1,230,000) quantum cascade laser mid-infrared imaging-guided MR-MSI dataset that enables isotopic fine structure analysis and therefore enhances the level of confidence substantially. To highlight the usefulness of the data, we compared 118 manual sulfatide annotations with the number of decoy database-controlled sulfatide annotations performed in Metaspace (67 at a false discovery rate <10%).

Overall, our datasets can be used to benchmark annotation algorithms, validate spatial biomarker discovery pipelines, and serve as a reference for future studies that explore sulfatide metabolism and its spatial regulation.

Figures