Bioinformatics Seminar
Time: 11AM
Venue: Davis Auditorium and Online
23 June 2026
Missing values and annotation in spatial metabolomics data
Tianyao LuWEHI
Spatial metabolomics can map biochemical heterogeneity directly in tissue, but two major data-quality issues currently limit interpretation. Missing values are extensive and non-random. Detection was intensity- and spatially dependent, with MNAR(Missing Not At Random) latent abundance proxies improving held-out detection prediction over technical-only models. This means absence cannot be treated as simple noise or zero abundance; it may encode ion-specific abundance, tissue structure, and acquisition effects. Besides, metabolite annotation remains ambiguous. The annotation workflow produced hundreds of candidate assignments. Even high confidence overlap peaks can map to multiple metabolite/adduct explanations. These analyses should be considered thoroughly such that we can address the analytical barrier of spatial metabolomics, rather than treating peak matrices as complete, directly identified metabolite maps.