Designing a Comparative Proteomics Experiment: Retention-time Alignment and Imputation Algorithms Affect Statistical Comparisons Between Samples

Comparative proteomics experiments reveal biomarkers by using statistical tests to determine proteins expressed with higher abundance in one sample versus another. However, comparative experiments can be complicated by variability from all aspects of proteomics workflows. To account for variability, software for database searching contains retention-time alignment and imputation algorithms to correct for retention-time shifts and assign abundances to missing proteins. While these algorithms improve quantification and reduce processing time, we hypothesize that they alter statistical comparisons between samples when samples are searched together. Herein, we search different cleanup methods or single proteins separately versus together in Progenesis Qi for proteomics database searching software. Our results show that searching samples together increases the number of identifications by each sample, enhances protein similarity between samples, and leads to false transfers. Further, we demonstrate that searching samples together affects protein abundance, differentially expressed proteins, and confidence scores due to retention-time alignment and imputation algorithms. Ultimately, we highlight that careful consideration of the search design is necessary to determine biomarkers in comparative proteomics experiments. Search results from the reanalyzed dataset comparing sample-cleanup methods (MSV000094130) and single-protein data have been deposited into MassIVE (MSV000096112) and ProteomeXchange (PXD056868).