Informatics Colloquium 10.12. Pushing the limits in automated NMR structure determination
Informatics Colloquium 10.12. 2019, 14:00 lecture hall D2
doc. RNDr. Tomáš Brázdil, Ph.D., FI MU
Pushing the limits in automated NMR structure determination
Abstract: Nuclear Magnetic Resonance (NMR) spectroscopy is a technique for
determining the structural and dynamic properties of molecules. NMR relies on
properties of certain atomic nuclei that, when placed in a strong magnetic
field, resonate at characteristic frequencies. NMR spectroscopy generates
complex data whose analysis is a laborious task and the road to NMR structure is
generally a long one. The chief barrier is that existing methodologies require a
significant amount of spectrometer time (several weeks), and effort by a trained
expert (up to several months). So fully automatic methods for NMR structure
determination are in demand.
In my talk, I will concentrate on the problem of assigning given resonant
frequencies (chemical shifts) from NMR spectra to the nuclei of a protein. Our
attempts at solving this demanding problem comprise both machine learning as
well as elementary graph theoretic methods. I will start with the machine
learning problem of predicting an amino acid type based on a given set of
chemical shifts. Consequently, I will show how such a prediction can be
incorporated into an algorithm for complete assignment of chemical shifts to
atoms in a protein using, among other methods, simple graph algorithms. I will
also present open problems and deficiencies of our current method which is still
under development.