Human Proteome Project Mass Spectrometry Data Interpretation Guidelines 2.1

Abstract

Every data-rich community research effort requires a clear plan for ensuring the quality of the data interpretation and comparability of analyses. To address this need within the Human Proteome Project (HPP) of the Human Proteome Organization (HUPO), we have developed through broad consultation a set of mass spectrometry data interpretation guidelines that should be applied to all HPP data contributions. For submission of manuscripts reporting HPP protein identification results, the guidelines are presented as a one-page checklist containing 15 essential points followed by two pages of expanded description of each. Here we present an overview of the guidelines and provide an in-depth description of each of the 15 elements to facilitate understanding of the intentions and rationale behind the guidelines, for both authors and reviewers. Broadly, these guidelines provide specific directions regarding how HPP data are to be submitted to mass spectrometry data repositories, how error analysis should be presented, and how detection of novel proteins should be supported with additional confirmatory evidence. These guidelines, developed by the HPP community, are presented to the broader scientific community for further discussion.

Keywords: Human Proteome Project; alternative protein matches; false-discovery rates; guidelines; mass spectrometry; standards.

Figure 1. Three scenarios for how false discovery rates combine

True positives are shown in blue, and false positives in red. The red false positive boxes are depicted approximately 10 times larger than they should be for enhanced readability. A) If none of the true positives and 1% false positives overlap, then the final FDR does not expand. B) If all of the true positives overlap, but none of the false positives overlap (because they are false and random), then the final FDR is double the original rates. C) In a real-world scenario where the intersection of true positives overlaps by 50% and the false positives do not overlap, the combined FDR is 1.5%. This effect compounds as more datasets are merged.