MSACL 2025 Abstract

A Procedural Approach to Data-Independent Acquisition-Based Amyloid Typing

William Phipps, Alexandra Emmanuel, Kelly Smith, Nik Krumm, Deanna Plubell, Han-Yin Yang, Christine Wu, McWilliam Catungal, Erinn Kuest, William Noble, Brian Searle, Michael MacCoss, and Andrew Hoofnagle
University of Washington

William Phipps, MD (Presenter) University of Washington

Presenter Bio: Bill Phipps is an Assistant Professor at the University of Washington in the Department of Laboratory Medicine and Pathology, where he serves as Assistant Director of Clinical Chemistry and is Clinical Pathology Residency Program director. He completed his residency training at UT Southwestern Medical Center in Dallas, TX.

Relevant Financial Disclosures (within past 24 months, reported on Apr 21, 2026) No relevant financial relationship(s) to disclose.

INTRODUCTION: Mass spectrometry remains underutilized in clinical laboratories for the analysis of solid tissues. Diagnostic workflows in anatomic pathology are still predominantly centered on immunohistochemistry (IHC), a widely adopted method. However, IHC presents significant challenges, including issues with standardization, reproducibility, quantitative precision, and dependency on monospecific antibody reagents and subjective visual interpretation. Furthermore, its limited capacity for multiplexing reduces its effectiveness in complex diagnostic scenarios. Mass spectrometry, known for its strengths in multiplexed protein analysis and reproducibility, offers a promising complementary approach for overcoming these challenges. For instance, data-dependent acquisition (DDA) mass spectrometry has become the gold-standard technique for tissue amyloid typing, enabling accurate identification of amyloid subtypes to guide targeted therapies. Meanwhile, data-independent acquisition (DIA) mass spectrometry has not yet been adopted in pathology services, reflecting a lack of standardized strategies for implementation, diagnostic interpretation, and validation.

METHODS: To address these gaps, we implemented a DIA-based pipeline for clinical amyloid typing using a Python-scripted, transferrable process. This involved the procedural generation of an application-specific amino acid sequence database using freely available, open-access resources from UniProt.org. We used the Comet search engine to analyze DDA results obtained from 325 unique formalin-fixed paraffin-embedded (FFPE) tissue blocks, representing 20 amyloid types (e.g., AA, Aβ, Aβ2M, ACal, AFib, AIns, AKer, ALκ, ALλ, ALys, ALECT2, AHλ, ATTR, and more) across 22 human tissue sources (e.g., bone marrow, heart, kidney, pancreas, and spleen). This DDA-based dataset covered ~600 LC-MS/MS injections and informed the construction of peptide sequence database for use in analyzing DIA results. In parallel, we conducted DIA LC-MS/MS analysis on an overlapping cohort of 175 cases of amyloidosis, covering 17 amyloid types and 21 tissue sources. We utilized Skyline software to extract peptide peak areas from DIA spectra and trained machine learning classifiers, including random forest algorithms, to predict amyloid type based on quantitative peptide measurements.

RESULTS: Using DDA, we identified >10,000 unique peptide sequences from amyloid tissue digests. Incorporating DIA analysis, we developed a random forest-based classifier trained on quantitative DIA peptide measurements. The classifier demonstrated a high level of robustness and precision, achieving an overall accuracy of 96.88% compared to DDA across five prevalent amyloid types (ALκ, ALλ, ALECT2, SAA, and ATTR). These measures were even further improved via the application of particular quality assurance (QA) measures, such as specifying requirements for the detection and measurement of spiked Peptide Retention Time Calibration (PRTC) Mixture peptides, a practice which could be readily applied in a clinical practice. Specific metrics include perfect precision, recall, and F1-scores for certain amyloid types (e.g., ALECT2, SAA, and ATTR), confirming the feasibility of peptide-based classification for complex diagnostic settings.

CONCLUSION: Quantitative peptide-based classification provides a viable framework for implementing DIA mass spectrometry in clinical tissue typing applications. For pathologists new to clinical amyloid typing, DIA offers distinct advantages over DDA, including its quantitative precision, multiplexing capabilities, and simplified workflow. Unlike traditional spectral-counting approaches that rely on relative intensity-based ranking, our pipeline leverages quantitative peptide signals as input features for robust classification. The accuracy demonstrated by the DIA-based classifier underscores its potential for routine use in pathology services, facilitating improved diagnostic workflows for amyloidosis and similar disorders.