Plyushchenko, I. V.; Luzzatto-Knaan, T.

Untargeted LC-MS metabolomics datasets often contain large numbers of redundant and non-informative features arising from background contaminants, multiple ion forms, poorly integrated peaks, and other low-quality signals. These features complicate downstream analysis by inflating feature space, degrading molecular networks, impeding pathway analysis, and obscuring statistically meaningful changes. Here, we present MetaboCensoR, an input-versatile Shiny application and local R package for analyte-centric peak table filtering. The workflow integrates four complementary modules for blank filtering, redundant ion-species filtering, quality-control filtering, and peak-based filtering. MetaboCensoR also provides interactive threshold optimization, exportable annotation tables, and synchronized filtering of associated .mgf files. The approach was evaluated across three independent datasets covering plant extracts, human cell lines, and bacterial interactions. Across these case studies, data filtering reduced feature redundancy and improved downstream interpretation in feature-based molecular networking, pathway-level functional analysis, and differential abundance testing, while preserving known target metabolites. These results show that systematic peak table filtering can substantially improve the interpretability and analytical value of untargeted metabolomics data.