Computational Biology and Language - Ambient Intelligence for Scientific Discovery - page 43

4.3 Prediction of Protein Folding Initiation Sites

The demonstrated success of language technologies for a number of typical com-
putational biology tasks suggests that these methods may also prove useful in
studies of tasks that have been studied less extensively. Prediction of folding
initiation sites in proteins is a formidable task that requires novel approaches.
We investigated if inverse frequencies may correlate with experimentally deter-
mined folding initiation sites in the protein folding model system, lysozyme. Our
hypothesis was based on the observation that in natural languages, rare words
carry the most relevant meaning of a text. Shown in Fig. 13 are inverse tri-gram
frequencies plotted along the lysozyme sequence. Indeed, we observed a correla-
tion between the locations of rare trigrams and the location of residual structure
in the unfolded protein as evidenced by maxima in relaxation rates measured in
NMR spectroscopic experiments. The statistical significance of this observation
remains to be established by extension to other proteins, but lysozyme is the only
protein for which the locations of folding initiation sites are known. However,
the steady growth in the size of the protein databank will allow a systematic
comparison between the sequences of n-grams and the number of structures that
each n-gram can occur in. Such statistics are already beginning to be reported in
the I-sites database [59] and in the analysis of sequences encoding certain types
of structures [60]. Thus, it is expected that n-gram analysis may significantly
contribute to the protein tertiary structure prediction problem in the future.

Fig. 13. Location of folding initiation sites in model protein Lysozyme (see Fig. 1) A.
Transverse relaxation rates [58]. Large values above the black line indicate the presence
of residual structure. B. Inverse trigram frequency in human lysozyme. The figure is
reproduced from [61] with permission from the publisher.