Abstract Background Mass spectrometry (MS) remains underutilized in the evaluation of formalin-fixed, paraffin-embedded (FFPE) tissue. Clinical deployment of MS in anatomic pathology is largely confined to data-dependent acquisition (DDA) work flows for amyloid typing. Alternative acquisition strategies such as data-independent acquisition (DIA) enable quantitative, peptide-centric tissue analysis and can expand MS use in biopsy evaluation. Methods We performed DIA-based LC-MS/MS analyses of amyloid-positive tissue specimens alongside an established DDA clinical work flow. The DIA pipeline incorporated automated data processing to support standardization and transferability. Peptide peak areas were extracted using Skyline, and machine-learning classifiers and peptide scoring rules were developed to predict amyloid type. Results DIA peptide measurements were evaluated across 304 amyloid-positive tissue specimens representing 16 discrete amyloid types. In a development group (n = 259), a random forest classifier enabled accurate classification of the five most common amyloid types from individual LC-DIA-MS/MS injections: λ and κ immunoglobulin light chain, transthyretin, leukocyte chemotactic factor 2, and serum amyloid A. In prospective evaluation (n = 45), the combined DIA classification strategy was fully concordant with the reference DDA clinical assignments. Amyloid types not represented in classifier training were resolved using peptide-level diagnostic signals evaluated against empirically derived intensity thresholds. Conclusions DIA-based mass spectrometry enables reproducible peptide-level classification of amyloid deposits in FFPE tissue. Combined with standardized data processing, machine-learning classification, and rule-based peptide scoring, this approach provides a scalable framework for proteomic amyloid typing. More broadly, it may support wider adoption of mass spectrometry for peptide biomarker quantification in anatomic pathology.
