Genotyping using ProQuant – you can but you shouldn’t!

One of the fantastic resources we have here at RxCelerate is our tissue BioBank. One set of samples contained within it was a set of longitudinal serum and plasma samples taken from healthy volunteers. Those samples were collected with different gradually decreasing frequency over a period of two years, and enabled us to investigate changes in biomarkers with time.

We analysed a subset of these samples (5 samples taken from each of 6 subjects over a month) using ProQuant. The analytical approach we used was one that we have developed over a period of years. Rather than focussing only on protein-level differences we use a non-hypothesis approach that includes open modification searches that enable us to identify any of the post-translational modifications included in the Unimod database. We therefore output datasets that include not only protein- and peptide-level information but also PTM and cleavage fractions.

While one has to interpret the data from such an open search carefully, it has the opportunity to find results you were not anticipating. In this case, because the Unimod database includes all 360 substitutions of one amino acid for any other as potential PTMs it has enabled us to find and validate multiple SNPs at the protein-level. As one can see from the following graphs, a series of amino acid substitutions were identified as PTMs, which were found to correspond to previously identified single nucleotide polymorphisms or SNPs. Few of these SNPs have previously been confirmed at the protein level, which means that for many this is the first documented evidence that proteins containing the altered sequences are expressed in vivo.

Clearly when measuring SNPs running longitudinal samples from the same individual is of little relevance. However, the data shown reinforces the precision of ProQuant, with every boxplot reflecting all of the measurement error from sampling, processing, LC-MS/MS and data processing of at least two peptides (with and without the SNP).

Note as well the data shows (as expected) that some peptides give different ‘response levels’ at the MS detector, meaning that heterozygotes are not always averaging to a PTM fraction of 0.5. Nevertheless, across all of the SNPs we identified using ProQuant the PTM fractions for homozygotes and heterozygotes were (mean with 95% CI):

0.020 (0.004 – 0.035)
0.452 (0.371 – 0.532)
0.938 (0.872 – 1.004) 

which only illustrates the quantitative performance of ProQuant yet again!

We are not, of course, recommending that ProQuant is used for genotyping individuals, although in this study we happened to validate expression of multiple SNPs in vivo. But we are strong advocates of approaching large datasets with an open, curious mind – never restrict observations to only those parameters you think in advance are of interest!