Benchmarking Peptide Detection Discrepancies Between DIA and DDA Using MS1 Trace  Analysis

Parisa Pirani

Mentors: Dr. Tytus Mak (Statistician), Dr. Meghan Burke Harris (Research Chemist), NIST Mass Spectrometry Data Center

Date/Time: April 29th, 2025 at 12:00pm

Abstract: Data-Independent Acquisition (DIA) and Data-Dependent Acquisition (DDA) are widely used in  proteomics, yet systematic biases in peptide detection between these methods remain an area of  active investigation. In this study, we analyzed publicly available raw data (PRIDE Archive:  PXD041421) from A549 human lung carcinoma cells, comprising three biological replicates and  four technical replicates per acquisition mode. DDA data were processed using FragPipe, which  incorporates MSFragger for peptide identification and IonQuant for quantification, while DIA data  were processed with DIA-NN in library-free mode. 

MS1 trace-based evidence was used to investigate acquisition-dependent dropout patterns by  examining peptides identified in both workflows for intensity concordance. A subset of shared  peptides exhibited zero intensity in DDA but quantifiable signal in DIA, despite meeting MS1- level requirements such as Gaussian peak shape, sufficient isotopic coverage, and reproducibility  across replicates. 

Further analysis of MS1 peak morphology and MS2 fragmentation characteristics was performed  using Skyline and IsoPique, a custom DDA-focused tool developed in-house at the NIST Mass  Spectrometry Data Center. The results revealed that many peptides with zero intensity in DDA  were supported by robust MS1 evidence, such as Gaussian peak shape, reproducibility, and  isotopic patterns, but did not meet IonQuant’s quantification thresholds, particularly the  requirement for a minimum number of isotopic peaks. Additionally, these peptides were often  associated with sparse or low-quality MS2 spectra, which are essential for confident identification  and quantification in DDA workflows. Conversely, Skyline’s quantification captured MS1 traces  for these peptides, although in some cases the traces may extend into regions not exclusively  associated with the identified precursor, raising the possibility of false positives in less restrictive  quantification settings. DIA, by design, integrates multi-fragment ion signals across retention time,  enabling detection of these peptides even when spectral similarity scores fall below typical DDA  thresholds. 

Overall, our findings underscore that discrepancies between DIA and DDA quantification are often  rooted not in detection failure, but in fundamental differences in fragmentation strategies,  quantification models, and filtering parameters. Integrating MS1-level evidence across workflows  provides a deeper understanding of these differences and can guide more informed interpretation  of peptide-level results in large-scale proteomics experiments.