Abstract
In this thesis, my work involving dierent aspects of protein structure determination by computer
modeling is presented.
Determination of several protein's native fold were carried out with Markov chain Monte Carlo
simulations in the PHAISTOS protein structure simulation framework, utilizing experimental
data in the form of chemical shifts, as well as distance restraints obtained either experimentally
or from sequence co-evolution. Of notable results, One of the determined structures, aKMT,
was not solved experimentally at the time, but was found to match the recently published X-ray
structure to within 3 A.
Furthermore, a fast quantum mechanics based chemical shift predictor was developed together
with methodology for using chemical shifts in structure simulations. The developed predictor
was used for renement of several protein structures and for reducing the computational cost
of quantum mechanics / molecular mechanics (QM/MM) computations of chemical shieldings.
Several improvements to the predictor is ongoing, where among other things, kernel based machine
learning techniques have successfully been used to improve the quantum mechanical level of theory
used in the predictions.
modeling is presented.
Determination of several protein's native fold were carried out with Markov chain Monte Carlo
simulations in the PHAISTOS protein structure simulation framework, utilizing experimental
data in the form of chemical shifts, as well as distance restraints obtained either experimentally
or from sequence co-evolution. Of notable results, One of the determined structures, aKMT,
was not solved experimentally at the time, but was found to match the recently published X-ray
structure to within 3 A.
Furthermore, a fast quantum mechanics based chemical shift predictor was developed together
with methodology for using chemical shifts in structure simulations. The developed predictor
was used for renement of several protein structures and for reducing the computational cost
of quantum mechanics / molecular mechanics (QM/MM) computations of chemical shieldings.
Several improvements to the predictor is ongoing, where among other things, kernel based machine
learning techniques have successfully been used to improve the quantum mechanical level of theory
used in the predictions.
| Originalsprog | Engelsk |
|---|
| Forlag | Department of Chemistry, Faculty of Science, University of Copenhagen |
|---|---|
| Antal sider | 158 |
| Status | Udgivet - 2016 |