Hidden molecular relationships are revealed by bootstrap resampling of mass spectral pairs with SpecReBoot

Abstract Mass spectral molecular networking organizes tandem mass spectrometry (MS/MS) data by connecting spectra based on similarity scores. However, these deterministic metrics provide no measure of uncertainty, so molecular networks often retain edges arising from noise or missing fragments, while authentic chemical relationships may remain obscured. To address this gap, we present SpecReBoot, a statistical framework that adapts Felsenstein’s bootstrap principle to MS/MS similarity scoring. By resampling fragment-level features and recomputing similarity across replicates, SpecReBoot transforms a single score into a confidence distribution and assigns bootstrap support to network edges. Using large, curated MS/MS spectral datasets, we demonstrate that SpecReBoot systematically removes unreliable connections and recovers robust low-similarity links, refining networks and revealing hidden relationships. We highlight how this confidence-aware “rebooting” guided the discovery of an unprecedented macrolactone scaffold from the endophytic fungus Diaporthe caliensis . Altogether, SpecReBoot provides the first general framework for quantifying confidence in MS/MS similarity and molecular network analysis.