This month, two particularly useful and imaginative papers were published in the proteomics world. The first from Neil Kelleher’s lab, (Molecular & Cellular Proteomics 15: 10.1074/ mcp.M114.047480, 45–56, 2016.) using an integrative approach with top-down and bottom-up proteomics and RNA sequencing to describe a model of breast cancer. The second was published by Claus Jorgense’s lab, (Tape et al., 2016, Cell 165, 1–11 May 5, 2016.) and uses an integrative global proteomics with multivariate phosphoproteomics to describe an oncogenic KRAS mutation effect on cellular signaling inside tumor cells and the tumor’s external stromal cells. Here’s our take on the first paper, hopefully, we’ll get time to write a summary of the second in short order.
Ioana Ntai et al., set out the large goal to compare the ability of top-down (TD) and bottom-up (BU) proteomics approaches to characterize different proteoforms that include post-translational modifications (PTMs), single nucleotide polymorphisms (SNPs) and novel splice junctions (NSJs) informed from transcripts. They completed three different experiments to determine these ends. Interestingly, any proteoform sequence that was consistent between human and mouse was removed from summary counts.
Not surprisingly, BU proteomics out performs TD for greater number of proteins identified. BU also identified more NSJs and SNPs as confirmed with transcripts. However, they show a nice example of where TD identified SNPs and protein products from heterozygous alleles. Most interestingly, in TD approach, closely related proteoforms, such as those with a SNP, co-elute in to the mass spectrometer and thus can be quantified pretty precisely. This allows relative quantification confirmation of heterozygous alleles. Although, TD approach is currently limited to lower molecular weight proteoforms, still over 1,000 proteoforms were able to be quantified with low FDR values between the two compared PDX model systems. The trend of greater spreads in fold change estimates coupled with higher confidence of their differential expression with TD is exciting. It begs to ask, how much relative quantification is removed due to inference of peptide quantification.
Comforting, is the fact that many of the relative quantification for the intersection of detection in both methods, is in agreement regarding relative abundance. Disagreements often arise from changes in PTM stoichiometry, so they say. Speaking of which, the TD approach really shines at detecting multiple isoforms and complex stoichiometry of PTMs that are completely lost with BU approaches. Finally, the authors document with all seriousness the effect of how utilizing less stringent thresholds for identification statistics has on the entire study’s quantification confidence. We see this effect all the time, but it’s SO very great to see it in action in the academic world with examples.