Differential Proteomics Analysis — Comparing Protein Abundance Across Conditions

Label-based methods (SILAC, TMT, iTRAQ) introduce stable isotope tags that allow multiple samples to be mixed and measured in the same MS run, reducing run-to-run technical variation and enabling direct intensity ratios. Label-free approaches quantify each sample in a separate run and compare intensities computationally; they require no specialised reagents and scale easily to large cohorts but demand strict chromatographic reproducibility and robust normalisation to control batch effects.

A minimum of three biological replicates per condition is required for any statistical inference; four to six replicates are strongly recommended for discovery experiments with moderate expected effect sizes. With only two replicates per group, variance cannot be reliably estimated and moderated tests offer little advantage. Power calculation tools such as PANDA or ProPower can help determine the sample size needed to detect a target fold change at a given FDR.

Proteins present in all replicates of one condition but absent from all replicates of the other ('all-or-nothing' proteins) represent a biologically important class. They should not be discarded. A common approach is to impute missing values from a low-intensity distribution (e.g., random draws from the 5th percentile of detected intensities) before statistical testing, and to report these proteins separately with clear annotation that they were subject to imputation.

The limma moderated t-test (originally developed for microarray data) is the most widely recommended approach for small-to-moderate sample sizes because it borrows variance information across all proteins to stabilise per-protein estimates. For larger datasets (n > 10 per group) ordinary t-tests or Wilcoxon tests perform comparably. DEqMS extends limma by weighting variance estimates by the number of peptides used for quantification, which can improve accuracy. Always apply Benjamini-Hochberg FDR correction for multiple comparisons.

Yes — multi-omics integration is a major application. After identifying differentially expressed proteins and differentially expressed genes separately, overlap and correlation analyses reveal whether changes are driven at the transcriptional or post-transcriptional level. Tools such as OmicsIntegrator, MOFA, or simple Pearson correlation between matched mRNA and protein fold changes are commonly used. Proteins that change without corresponding mRNA changes are candidates for regulation by translation efficiency, protein stability, or post-translational modifications.