Flagging false positives following untargeted LCMS characterization of histone PTM combinations

Epigenetic changes can be studied with an untargeted characterization of histone post-translational modifications (PTMs) by LCMS. While prior information about more than twenty types of histone PTMs exists, little is known about histone PTM combinations (PTMCs). Due to the combinatorial explosion it is intrinsically impossible to consider all potential PTMCs in a database search. Consequentially, high-scoring false positives with unconsidered but correct alternative isobaric PTMCs can occur. Current quality controls can neither estimate the amount of unconsidered alternatives nor flag potential false positives. Here, we propose a conceptual workflow that provides such options. In this workflow, an in silico modelling of all candidate isoforms with known-to-exist PTMs is made. The most frequently occurring PTM sets of these candidate isoforms are determined and used in several database searches. This suppresses the combinatorial explosion while considering as many candidate isoforms as possible. Finally, annotations can be classified as unique or ambiguous, the latter implying false positives. This workflow was evaluated on an LCMS dataset containing 44 histone extracts. We were able to consider 60% of all candidate isoforms. Importantly, 40% of all annotations were classified as ambiguous. This highlights the need for a more thorough evaluation of modified peptide annotations