Poster Presentation AUS-oMicS 2025

Benchmarking imputation methods, database search engines, and mass spectrometry instrumentation: applications to quantitative proteomics studies (119924)

Jumana Yousef 1 2 , Vineet Vaibhav 1 2 , Che Stafford 1 2 , Sukhdeep Spall 1 2 , Laura F Dagley 1 2
  1. The Walter and Eliza Hall Institute of Medical Research, Parkville, VIC, Australia
  2. Department of Medical Biology, University of Melbourne, Parkville, VIC, Australia

Background: In mass spectrometry-based proteomics studies, accurate protein quantification is critical for reliable biological interpretations. Variations introduced by different search engines, mass spectrometry instruments, and imputation methods can significantly impact data quality and quantitative accuracy.

Aim: This study compares the effects of various imputation strategies (including Perseus, msImpute, and limpa, versus no imputation), search engines (MaxQuant, DIA-NN, and Spectronaut), and mass spectrometry instruments (Orbitrap Eclipse vs. Orbitrap Astral) on protein quantification reliability and accuracy. The research aim was to identify the total number of proteins present in human cancer tissues. The specific protein class are enzymes that are known to be present in low abundance.

Methods: Whole cell lysates were prepared for mass spectrometry analysis using the USP3 digestion method. Peptides were analysed on both Oribitrap Eclipse and Orbitrap Astral MS instruments Using a 15 cm IonOpticks column which data acquired in a data independent acquisition (DIA). Resulting MS data was analysed with DIA-NN and Spectronaut. Datasets were evaluated for peptide identification consistency, quantification accuracy, protein abundance variability, and the ranking of positive controls, with and without imputation.

Results: Preliminary results show significant differences in peptide identification rates, with the Astral MS displaying higher sensitivity. DIA-NN offered the best balance of identification rates and computational efficiency. Imputation generally increased apparent protein abundances, introducing higher variability, notably in MaxQuant processed datasets. Combining Astral with Spectronaut and subsequent imputation with limpa resulted in the lowest variability across replicates.

Conclusion: The choice of MS instrument, database search engine, and imputation strategy profoundly affects proteomics data quality. While imputation facilitates more comprehensive analyses, its impact on variability requires careful management. The integration of specific instruments and software is crucial for maintaining data integrity.