Significance analysis of microarray for relative quantitation of LC/MS data in proteomics

BACKGROUND: Although fold change is a commonly used criterion in quantitative proteomics for differentiating regulated proteins, it does not provide an estimation of false positive and false negative rates that is often desirable in a large-scale quantitative proteomic analysis. We explore the possibility of applying the Significance Analysis of Microarray (SAM) method (PNAS 98:5116-5121) to a differential proteomics problem of two samples with replicates. The quantitative proteomic analysis was carried out with nanoliquid chromatography/linear iron trap-Fourier transform mass spectrometry. The biological sample model included two Mycobacterium smegmatis unlabeled cell cultures grown at pH 5 and pH 7. The objective was to compare the protein relative abundance between the two unlabeled cell cultures, with an emphasis on significance analysis of protein differential expression using the SAM method. Results using the SAM method are compared with those obtained by fold change and the conventional t-test. RESULTS: We have applied the SAM method to solve the two-sample significance analysis problem in liquid chromatography/mass spectrometry (LC/MS) based quantitative proteomics. We grew the pH5 and pH7 unlabelled cell cultures in triplicate resulting in 6 biological replicates. Each biological replicate was mixed with a common (15)N-labeled reference culture cells for normalization prior to SDS/PAGE fractionation and LC/MS analysis. For each biological replicate, one center SDS/PAGE gel fraction was selected for triplicate LC/MS analysis. There were 121 proteins quantified in at least 5 of the 6 biological replicates. Of these 121 proteins, 106 were significant in differential expression by the t-test (p < 0.05) based on peptide-level replicates, 54 were significant in differential expression by SAM with Δ = 0.68 cutoff and false positive rate at 5%, and 29 were significant in differential expression by the t-test (p < 0.05) based on protein-level replicates. The results indicate that SAM appears to overcome the false positives one encounters using the peptide-based t-test while allowing for identification of a greater number of differentially expressed proteins than the protein-based t-test. CONCLUSION: We demonstrate that the SAM method can be adapted for effective significance analysis of proteomic data. It provides much richer information about the protein differential expression profiles and is particularly useful in the estimation of false discovery rates and miss rates.