In the course of the statistical analysis of protein (peptide) fragmentation due to mass spectrometry, the framework of _wHMMs_ (weighted hidden markov models) has been introduced. The first step of my PhD-project concerns the extension of the underlying sequence model (from i.i.d. to Markovian).
Attending the fact that the amino acid positions are not independent, but rather determine the conformation and function of a protein via secondary and tertiary structure, I used an approx. model selection criterion, namely the Bayesian information criterion (BIC), to find out whether a markovian model affords a more plausible (reasonable) description of real protein (peptide) sequences.
Finally, a resulting wHMM as well as ideas for future work shall be discussed.