Giovanni Minervini
Department of Biology, University Roma Tre, Rome, Italy

Massive Non Natural Proteins Structure Prediction Using Grid Technologies

Motivation: The number of natural proteins is an infinitesimal fraction of
all the theoretically possible protein sequences. In fact, considering
random protein sequences of only 100 amino acids it is possible to obtain
100^20 structurally different proteins. Thus, there is an enormous number
of proteins never exploited by nature or, in other words, "never born
proteins (NBPs)". A fundamental question in this regard is if the ensemble
of natural proteins possesses peculiar properties in terms for example of
thermodynamic, kinetic or functional properties. A key feature of natural
proteins is the ability to form a stable and well defined three dimensional
structure. Thus, the structural study of NBPs can help to understand if
natural protein sequences were selected during molecular evolution for their
peculiar properties or if they are just the product of contingency.
This problem cannot be approached experimentally, as this would require the
structural characterization of a huge number of random proteins.
Thus we chose to tackle the problem using a computational approach.

Methods & Results: A random protein sequences library (2x10^4 sequences)
was generated using the utility RandomBlast which produces random amino
acid sequences with no significant similarity to natural proteins.
The structural properties of NBPs were studied using the ab initio protein
structure prediction software Rosetta (Rohl et al. Methods Enzymol. 2004;
383, 66-93). Given the highly computational demanding problem,
the Rosetta software was ported in the EUChinaGRID infrastructure
(http://www.euchinagrid.org) and a user friendly job submission environment
was developed within the GENIUS Grid Portal (https://genius.ct.infn.it/).
Protein structures generated were analysed in terms of secondary structure
content, overall topology, surface/volume ratio, hydrophobic core composition,
net charge. Results obtained indicate that the vast majority of NBPs,
according to the Rosetta model, are characterized by a compact
three-dimensional structure with a high secondary structure content.
Structure compactness is comparable to that of natural proteins,
suggesting similar stability.
Deviations are observed in hydrophobic core composition, as NBPs appear
to be richer in aromatic amino acids with respect to natural proteins.
The results will be discussed in view of the evolutionary implications
of NBPs properties both at the amino acid and nucleotide level.