Post-translational modifications (PTMs) are molecular groups that are added to proteins after
translation. As is the case throughout most of proteomics research, the method of choice for their
analysis is tandem mass spectrometry. However, because the analysis of PTMs is not as
straightforward as the analysis of normal proteins, many PTMs remain undetected. One of the
most abundant and most studied PTMs is phosphorylation, which plays a key role in signal
transduction, and is disrupted in diseases as varied as Alzheimer’s and cancer. Because of this large
interest and broad application domain, my research will focus on improving the detection of
phosphorylation.
My objective is to create a novel method for the identification of phosphorylated proteins, based
on the re-use of the large amounts of data in public proteomics databases. Indeed, these data can
be used to train machine learning algorithms that can improve the identification of
phosphorylated proteins. This can occur because existing data can teach the machine learning
algorithms to predict an accurate, theoretical spectral library for all possible phosphorylated
peptides, complete with detailed peak intensity information. This should lead to an unbiased
search engine with a much higher sensitivity than existing methods. Moreover, this data-driven
approach can easily be adapted to all current, and possible future fragmentation techniques.