nontarget

nontarget detects homologous compound series in liquid chromatography–high-resolution mass spectrometry (LC-HRMS) data to enable untargeted identification of repeating-unit anthropogenic compounds by exploiting mass and retention-time patterns.

Key Features:

• Two-Staged Detection Approach: Employs a two-stage algorithm that uses a k-d tree representation of selected LC-HRMS peaks to identify feasible 3-tuples of peaks filtered by criteria such as mass-defect differences, then recombines these 3-tuples into larger series while enforcing smooth retention-time transitions.
• Unsupervised Methodology: Operates without prior compound annotations, enabling assignment of many nontargeted peaks into continuous homologous series.
• Resolution of Ambiguities: Addresses assignment ambiguities, including isobaric and gapped series peaks, using a self-organizing map technique to distinguish rival combinatorial solutions and series meshing.

Scientific Applications:

• Environmental Monitoring: Applied to effluent samples from Swiss sewage treatment plants (STPs) to identify ubiquitous and partially low-frequency series mass differences across STPs and to inform prioritization of identification efforts.
• Analytical Chemistry: Supports detection and analysis of complex chemical mixtures in LC-HRMS datasets to provide insights into the presence and behavior of homologous compounds.

Methodology:

The algorithm partitions LC-HRMS signal characteristics, uses a k-d tree and a nearest-neighbour walk to extract feasible 3-tuples of peaks filtered by mass-defect differences, recombines those triplets into larger series while enforcing smooth retention-time transitions, addresses ambiguities with a self-organizing map, and represents two mass-defect-related dimensions by a single dimension for illustration.