Evaluation of a Consensus Based MS/MS Library Matching Algorithm and Integration into a Non-Targeted Software Platform for Increasing Exposome Coverage

Significance of the topic

The reliable identification of per- and polyfluoroalkyl substances (PFAS) and co-occurring anthropogenic contaminants in complex environmental matrices is critical for exposome studies, regulatory screening, and remediation prioritization. MS1-based classification approaches (mass defect, homologous series, Kaufmann plots) are useful for defining chemical space but can produce false positives because unrelated compounds may occupy the same MS1 domains. Integrating MS/MS spectral evidence and robust multi-metric scoring into non-targeted workflows increases annotation confidence and expands detectable chemical coverage in environmental monitoring.

Objectives and overview of the study

This work implemented MS/MS library searching and a consensus-based scoring algorithm within the FluoroMatch non-targeted software platform and evaluated its impact on PFAS annotation and overall exposome coverage in construction and demolition landfill leachate. Specific goals were to (1) adapt MassBank and NIST DIMSpec libraries for rapid querying, (2) compare complementary MS/MS similarity metrics, and (3) demonstrate how consensus ranking corrects MS1-driven misassignments and enables discovery of non-PFAS co-contaminants.

Methodology

The workflow compares experimental MS/MS spectra for each detected feature to candidate reference spectra from two libraries (MassBank and NIST DIMSpec). Key elements:

• Library preparation: NIST DIMSpec used in native format; MassBank records restructured into a unified SQLite database for efficient searching.
• Candidate filtering: Precursor m/z tolerance (ppm) and ionization polarity applied before spectral comparisons.
• Spectral preprocessing: Alignment and removal of low-quality signals and noise prior to scoring.
• Similarity metrics: Three complementary scores calculated—dot-product, reverse cosine similarity, and spectral entropy similarity.
• Consensus ranking: Individual metric scores aggregated into a consensus-based rank to prioritize candidates and reduce metric-specific biases.
• User controls: Interactive filters to inspect and prioritize identifications by individual metric values, fragment occurrences, or the consensus rank; visualization of matched spectra for manual review.

Used instrumentation

The poster describes software and library integration rather than specific mass spectrometer models. Implemented components include:

• FluoroMatch non-targeted analysis platform (software integration).
• Reference libraries: MassBank (v2025.10, restructured into SQLite) and NIST DIMSpec library (native format).
• Scoring algorithms: dot-product, reverse cosine, and spectral entropy metrics implemented within the FluoroMatch environment.

No specific mass analyzer, ion source, or LC conditions were reported in the provided text.

Main results and discussion

Key findings and observations from applying the consensus MS/MS approach to landfill leachate samples:

• MS1-only classification limitations: Features that matched PFAS chemical-space in Kaufmann plots (mass defect vs carbon, mass vs carbon domains) and appeared in homologous series were sometimes non-PFAS, demonstrating false positives arising from MS1 criteria alone.
• MS/MS evidence resolves misclassifications: Example—4-isopropylbenzenesulfonic acid (C9, m/z 199.0434) was initially flagged as PFAS by MS1 metrics but MS/MS library matching favored non-PFAS spectra, allowing correction of the tentative assignment.
• Consensus scoring increases robustness: Individual metrics responded differently to spectral quality (e.g., noisy spectra produced conflicting metric values). Aggregating dot-product, reverse cosine, and entropy-based scores produced more consistent rankings than any single metric alone, improving reliability across variable spectra.
• Expanded exposome coverage: Incorporation of MassBank enabled identification of numerous non-PFAS co-contaminants commonly found in construction and demolition debris. Annotated compounds included phenolic antioxidants (e.g., Fenozan, 4,4'-Thiobis(6-tert-butyl-m-cresol)), organophosphate additives (e.g., tributyl phosphate), and linear alkylbenzene sulfonates (LAS).

Benefits and practical applications

The consensus MS/MS integration provides several practical advantages for environmental and non-targeted screening workflows:

• Reduced false positives in PFAS screening by providing orthogonal fragmentation evidence to MS1-based classification.
• Improved annotation confidence through multi-metric consensus ranking, particularly valuable for low-quality or noisy spectra.
• Broader chemical coverage by leveraging large public libraries (MassBank, NIST DIMSpec), enabling detection of co-contaminants relevant to exposure assessment and waste characterization.
• Interactive visualization and filtering tools support manual validation and prioritization, aiding QA/QC in laboratory workflows.

Future trends and potential uses

Opportunities to extend and refine the approach include:

• Expanded and curated spectral libraries: Continued community contribution and harmonization of library formats will increase identification rates for diverse chemistries.
• In silico spectral prediction and hybrid matching: Combining experimental libraries with predicted spectra to fill gaps for novel or proprietary compounds.
• Machine learning–driven consensus models: Data-driven weighting of individual similarity metrics or incorporation of orthogonal features (retention time, ion mobility) to improve ranking performance.
• Automated QA thresholds and provenance tracking: Standardized confidence scoring and traceable decision logic to support regulatory and long-term monitoring programs.
• Acquisition strategy optimization: Tailoring MS/MS acquisition (DDA/DIA settings, collision energies) to maximize spectral quality for consensus-based identification.

Conclusion

Integrating MS/MS library searches and a consensus-based scoring strategy into FluoroMatch demonstrably reduces MS1-driven misassignments and expands exposome coverage in landfill leachate. The use of complementary similarity metrics mitigates individual-method biases and increases resilience to variable spectral quality. Broad public libraries such as MassBank, combined with flexible visualization and filtering, enable identification of PFAS and numerous co-contaminants, improving the comprehensiveness and confidence of non-targeted environmental analyses.

References

1. MassBank (v2025.10). Zenodo dataset.
2. NIST DIMSpec Library. Journal of the American Society for Mass Spectrometry, 2024. DIMSpec library release.
3. Li, Y. et al. Spectral entropy outperforms MS/MS dot product similarity for small-molecule compound identification. Nature Methods 18, 1524–1531 (2021).