Abstract :
[en] De novo protein design is a growing field in protein chemistry, aiming at the production of
artificial proteins. On a purely fundamental basis, the design of proteins from scratch allows
testing the accuracy of the current protein knowledge and, possibly, to improve it. A deep
knowledge of the sequence, structure, function relationships in proteins is necessary to design
new proteins with specific functions. This facet of de novo protein design has numerous appli-
cations in biotechnology and biomedicine. On the other hand, in the context of a post-genomic
era, advanced computational methods for protein analysis, modelling and design are needed to
decode the massive amount of genomic data.
There is a long tradition at the University of Liège in the design of artificial (β/α) 8 -barrel
proteins, called Octarellins. This fold, also known as TIM-barrel, is widespread in nature, partic-
ularly in enzymes, and represents an interesting target for therapeutic or biological applications.
Several generations of Octarellins were designed with the help of very different approaches.
Lessons from these previous works has served as a rational basis for this study, which consists in
the design of a new generation of artificial TIM-barrels, termed OctaVII. This thesis is divided
in four sections that are shortly described hereafter:
The first section describes a pool of natural TIM-barrels, which structural features were
analyzed in order to extract useful information for the following steps of design and validation.
The second section is dedicated to the design of OctaVII models. Backbone structures were
designed with the use of the modelling software Rosetta and Modeller. This led to the selection
of 28 backbones structures, which were used for the design of sequences, using Rosetta. Finally,
more than 8000 artificial sequences were designed.
The third section includes the in silico validation of the design. Information obtained from
natural TIM-barrels was used to screen the 8000 artificial sequences and to select 10 of them for
experimental characterization. Various structural features were tested, including hydrogen bond
content and amino acid composition, and both secondary structure predictions and molecular
dynamic simulations were performed.
The fourth section is dedicated to the experimental validation of the design through pro-
tein expression, purification and biophysical characterization. In addition to the ten original
sequences that were designed in this work, five additional variants were tested for their possibly
improved properties (collaborations at the Vrije Universiteit Brussel, Belgium, and at the Van-
derbilt University, USA).
This thesis contributes to the development of the de novo design of proteins as an emerging
methodology for both a better understanding of proteins and the design of new functional
proteins with applications in biomedicine and nanotechnology.
