Morphological Clustering for AI-based Quality Control of LC–MS

May 20, 2026

Reliable peak detection and quality control are essential for LC‑MS peptide mapping, particularly in routine workflows that must be harmonized across instruments and organizations. Assessing peak quality and grouping isotopic signals consistently remains challenging, especially in high‑throughput environments where errors propagate into downstream analysis.

This poster presents a physically grounded, text‑conditioned clustering approach that applies LLM‑based grouping to LC‑MS peak data, combining Gaussian‑derived peak features with expert‑defined criteria to enable interpretable and auditable quality control. You will learn how this approach supports grouping of LC‑MS peaks based on morphology, signal strength, and detectability, reduces isotope clustering errors that impact downstream peptide mapping, and enables interpretable, rule‑driven QC without requiring supervised training. It also demonstrates how advanced AI methods can be applied within the end‑to‑end process traceability of the Genedata Expressionist^® platform, providing a foundation for trustworthy, transparent analysis rather than isolated, black‑box approaches. Designed for scientists and analysts working with LC‑MS peptide mapping workflows.