The reproducible acquisition of comparative liquid chromatography/tandem mass spectrometry data from complex biological samples

An in‐depth study of the reproducibility of data acquired for comparative proteomics analysis using a prototype two‐stage heated laminar flow chamber fitted to a commercial high‐performance liquid chromatography/tandem mass spectrometry (HPLC/MS/MS) instrument was undertaken. The study is based on 24 replicate samples from four independent membrane preparations derived from two matched breast cancer cell lines. Variation and reproducibility in the data were evaluated at several levels highlighting the relative efficiency and variability of the acquisition routines used. Specifically, variation in the number and relative intensities of chromatographic peaks eluted from the LC column, precursor ion selection and sequence identification were evaluated. On average, approximately 6500 chromatographic peaks were generated for each acquisition with a corresponding coefficient of variance (CV) of less than 20%. Precursor ion selection and sequence identification averaged 1380 and 780 events per acquisition sample, respectively, with corresponding CVs of less than 10% for each. The reproducibility in the precursor ion selection was typically better than 60% between similar replicates. Using protein and peptide internal standards, it was found that the CV in retention time across the gradient between two acquisition pairs was typically less than 5%, whereas the average intensity ratio was 1.0 (expected) with a CV approaching 20%. An evaluation of the intensity ratios calculated from endogenous peptide sequences, identified across the acquisition set, indicated a CV of ∼30%. Similarly, the CV associated with the top 1000 peptides indicated a mean and median of 28.4 and 26.95%. For a given acquisition pair it was also found that ∼11% of the chromatographic peaks eluting from the column were linked to a sequence or identified. For these experiments, less than 10% of the peak pairs had absolute ratios greater than 2.0 and of those only ∼10% had sequences linked to them. For each matched acquisition set on average 406 proteins were identified with a CV of less than 10%. Of the proteins that were identified approximately 30% had at least one predicted trans‐membrane domain, indicating a four‐fold increase over a crude homogenate sample with only minor enrichment. During these experiments it was found that the interface did not significantly alter the relative charge state distribution of ions, nor did it introduce significant interference from background ions. The interfa