Insight Profiler streamlines analysis of non-targeted metabolomics data from LCMS or direct injection in a single software solution

Importance of the topic

The development of robust, integrated data‑processing tools for non‑targeted metabolomics is essential to translate complex high‑resolution mass spectrometry (HRMS) data into biologically and clinically meaningful findings. Automating feature detection, alignment, statistical interrogation and spectral annotation in a single reproducible processing cascade reduces workflow fragmentation, improves throughput for large cohorts and supports discovery of candidate biomarkers for diseases such as pancreatic ductal adenocarcinoma (PDAC).

Objectives and study overview

This work presents Insight Profiler, a software application designed to perform end‑to‑end non‑targeted metabolomics data processing within a single configured method. The software was evaluated on a biomarker discovery study comparing serum from PDAC patients (n=30) and healthy controls (n=30). Samples were analyzed by two complementary HRMS approaches: reversed‑phase LC‑DIA‑MS/MS on a Q‑TOF and rapid direct probe ionization mass spectrometry (DPiMS) without chromatography. The principal goals were to demonstrate automated feature extraction through to compound annotation, and to compare results between LC‑MS/MS and direct injection workflows using the same processing cascade.

Materials and methods

• Sample set: serum extracts from PDAC patients (n=30) and healthy controls (n=30) obtained under ethical approval.
• LC separation: reversed‑phase C18 BEH column (2.1 × 100 mm, 1.7 µm) at 50 °C, 0.4 mL/min, binary gradient (water + 0.1% formic acid / acetonitrile + 0.1% formic acid), ~35 min cycle time.
• LC‑DIA‑MS/MS acquisition: TOF MS survey scans and data‑independent acquisition (DIA) MS/MS using 27 windows (m/z range indicated), polarity switching, collision energy spread 5–55 V; high‑resolution Q‑TOF detection (m/z range up to 1500).
• DPiMS acquisition: high‑resolution direct probe ionization Q‑TOF analysis for minimal‑preparation, high‑throughput profiling of the same serum extracts without LC separation; positive and negative ion features extracted simultaneously.
• Data processing: Insight Profiler configured as a single method cascade performing peak/feature detection, retention/time alignment (where applicable), statistical filtering (including PCA, volcano plots, boxplots), and compound annotation against an in‑house metabolomics library and external repositories (MassBank, LipidBlast, HMDB).

Used instrumentation

• Shimadzu LC system with C18 BEH column for reversed‑phase separations.
• Shimadzu LCMS‑9030 Q‑TOF (TOF MS and DIA‑MS/MS) for LC‑DIA‑MS/MS experiments.
• Direct Probe Ionization MS (DPiMS) source coupled to Q‑TOF for rapid, direct injection profiling.
• Insight Profiler software (Shimadzu Corporation) for unified data processing and visualization.

Main results and discussion

• Insight Profiler successfully automated the entire non‑targeted workflow—feature extraction, alignment, statistical comparison and tentative compound identification—capturing all steps in a single processing method suitable for batch analysis.
• LC‑DIA‑MS/MS analysis identified a set of statistically significant metabolite features that differentiated PDAC serum from controls. Notable observations included decreased levels of phospholipids enriched in linoleic acid (examples: LPC 18:2, PC 18:1_18:2, PC 18:2_18:2, PC 18:2_20:4) and reduced LPE 18:2, LPC 18:1, LPC 18:3 and LPC 20:5. Glutamic acid was found to be significantly elevated in PDAC samples. These annotations were supported by MS/MS fragmentation patterns and cross‑mode detection (positive and negative ions) to increase confidence in putative IDs.
• Applying the same Insight Profiler processing settings to DPiMS data reproduced many of the statistically significant features observed in LC‑MS/MS, including an ion (m/z 542.32218, putatively LPC 18:2) showing lower abundance in PDAC samples. This concordance demonstrates DPiMS as a potential rapid screening tool for candidate biomarkers discovered by LC‑MS/MS, despite the lack of chromatographic separation.
• Visualization and statistical outputs (PCA, volcano plots, boxplots) were linked to spectral and library matches, enabling efficient triage of candidate markers directly within the software environment.
• Limitations: identifications are putative and rely on library matches and MS/MS evidence; DPiMS lacks chromatographic separation and may be vulnerable to isobaric interferences and ion suppression, requiring follow‑up validation with targeted LC‑MS/MS, orthogonal techniques, and larger cohorts for clinical translation.

Benefits and practical applications

• Workflow simplification: a single configurable processing cascade reduces manual data transfers between tools, improving reproducibility and lowering operator burden for large‑scale metabolomics projects.
• High‑throughput screening: DPiMS coupled with the same processing pipeline enables rapid triage of samples for biomarker presence/absence, supporting screening or prioritization workflows where speed is critical.
• Integrated interpretation: linking statistical outputs directly to spectra and library hits accelerates candidate verification and reduces time from discovery to validation planning.
• Transferability: unified processing methods can facilitate consistent batch processing across multiple datasets and potentially across laboratories, if standardized libraries and parameters are adopted.

Future trends and opportunities

• Validation and clinical translation: follow‑up targeted assays, larger multi‑center cohorts and orthogonal validation (e.g., targeted LC‑MS/MS, immunoassays) will be needed to confirm biomarker utility and assess diagnostic performance for PDAC.
• Hybrid workflows: combining rapid DPiMS screening with confirmatory LC‑MS/MS or ion mobility separation to resolve isobaric species could produce efficient discovery→validation pipelines.
• Expanded libraries and AI‑assisted annotation: growth of curated spectral libraries and application of machine learning for spectral annotation and pattern recognition will improve identification confidence and speed.
• Standardization and reproducibility: community standards for processing parameters and reporting will enhance comparability between studies and support regulatory acceptance of biomarker panels.
• Integration with multi‑omics and clinical data: combining metabolomics outputs processed in a unified software environment with genomics, proteomics and clinical metadata will strengthen biomarker discovery and mechanistic interpretation.

Conclusion

Insight Profiler demonstrates that end‑to‑end automation of non‑targeted metabolomics—from feature detection and alignment to statistical filtering and spectral annotation—can be implemented in a single processing method. When applied to a PDAC serum study, the software identified biologically plausible metabolite alterations (notably changes in linoleic‑acid‑containing phospholipids and glutamic acid) and produced concordant results between LC‑DIA‑MS/MS and high‑throughput DPiMS. The approach reduces workflow complexity and indicates that rapid direct‑injection techniques, processed with the same rigorous pipeline, can serve as practical screening tools prior to targeted validation. Further validation, larger cohorts and refined identification workflows are required to advance these candidate markers toward clinical utility.

References

The original study used internal and public spectral resources (in‑house metabolomics library, MassBank, LipidBlast, HMDB) for annotation. No formal literature list was provided in the source material.