Homologous Series Detection

Homologous Series Detection detects homologous chemical series in hyphenated high-resolution mass spectrometry (LC-HRMS) data by nontargeted extraction of compound sets that differ by repeating chemical units to support analysis of polar anthropogenic environmental contaminants.

Key Features:

• Nontargeted Approach: Operates without prior compound-specific information to enable comprehensive detection across LC-HRMS datasets.
• Two-Stage Extraction Process: Uses a k-d tree representation of LC-HRMS peaks to identify feasible 3‑tuples based on mass defect differences in the first stage and recombines these 3‑tuples into larger series while enforcing smooth retention time transitions in the second stage.
• Partitioned Signal Representation: Employs a partitioned representation of LC-HRMS signal characteristics to simplify complex data dimensions for series extraction.
• Unsupervised Methodology: Functions without predefined models or training data, enabling application to novel datasets.
• Ambiguity Resolution: Applies a self-organizing map to resolve ambiguities from isobaric or gapped series peaks and to reveal complex series interactions.
• Evaluation and Validation: Was evaluated on ten effluent samples from Swiss sewage treatment plants (STPs) and demonstrated capability in both positive and negative electrospray-ionization modes.
• Ubiquitous Series Identification: Detects common yet low-frequency mass differences across STPs to support prioritization for further identification and research.

Scientific Applications:

• Environmental Monitoring: Enables detection and analysis of homologous compounds in environmental samples such as sewage treatment plant effluents.
• Chemical Analysis: Supports comprehensive chemical profiling by identifying diverse LC-HRMS peak series that may be missed in targeted analyses.
• Research Prioritization: Highlights prevalent mass differences across datasets to guide subsequent identification and follow-up studies.

Methodology:

The method uses a partitioned representation of LC-HRMS signal characteristics and a k-d tree representation of peaks to extract feasible 3‑tuples via a nearest-neighbour walk based on mass defect criteria; extracted 3‑tuples are recombined into larger series enforcing smooth retention time transitions, and a self-organizing map is used to resolve ambiguities from isobaric or gapped peaks.