The transmembrane proteins are integral membrane proteins with one or more segments that span the entire biological membrane. They are essential in maintaining the normal cell physiology, playing vital roles in ligand transport, cell signaling, energy production, immunity etc. Besides, the transmembrane proteins are important drug targets and are responsible for their carrier-mediated active uptake. A rough estimate suggests around 25% of the sequenced genome to code for transmembrane proteins. Nevertheless, the present knowledge about the transmembrane protein sequences and their classification remains incomplete. Moreover, due to experimental difficulties, the structural and functional details of the known transmembrane proteins are vastly underexplored. Considering the biological and pharmaceutical interest, several in-silico methods are developed and applied, in combination with experimental methods, towards elucidating the structural details and functional mechanisms of the transmembrane proteins. The research work presented in this thesis is focused on developing transmembrane prediction algorithms. Identifying the transmembrane regions is the preliminary step towards detailed structural understanding of a transmembrane protein. The developed algorithms show two-layer architecture. The first layer consists of a data-driven classifier built using Support Vector Machines. The classification is based on the amino acid patterns of the transmembrane regions are characterized mathematically using amino acid adjacency matrix. In the second layer, the inital prediction is refined using the statistical data on amino acid preference patterns at at the trnasmembrane region boundaries. The algorithms are trained on data that provide a better representation of the available transmembrane protein space. Compared to other available predictors, the developed algorithms show higher prediction accuracy and better performance, especially in case of beta transmembrane proteins. Being independent of evolutionary information derived from sequence profiles is another advantage of the developed algorithms, which are able to provide transmembrane region predictions in case of novel non-homologous sequences with potentially higher accuracy. The next part of the thesis is directed towards the structural elucidation of bilitranslocase, an organic anion transporter protein. It is a 340 amino acids long transmembrane protein with no available structural information. To identify the transmembrane regions of bilitranslocase, the developed transmembrane region prediction algorithm is used. Later, the stability of the predicted transmembrane domains are assessed using molecular dynamics simulations. The probable arrangements of transmembrane domains are studied, which identifies the most populated and energetically favourable arrangement types. NMR studies are performed to obtain the 3D structures of the transmembrane domains playing central role in the transport mechanism. The work analyzes and discusses in details the possible functional importance of the bilitranslocase structural details obtained from the theoretical and NMR experiments.
D.09 Tutoring for postgraduate students
COBISS.SI-ID: 268364288The subject of invention is bitumen with the addition of PMMA/ATH composite dust. PMMA/ATH composite powder is, according to the invention, added to the bitumen, which is used in asphalt mixtures, in order to improve properties of asphalt mixtures. The invention belongs to the field of asphalt. Modified bitumen, according to the invention contains the following ingredients in percentage by weight: from 50% to 99.9% (w/w) binder, which is mostly bitumen, and from 0.1 to 50% (w/ ) PMMA / ATH composite powder. Bitumen properties were tested with standard tests, such as softening point test method ring&ball (EN 1427), penetration test (EN 1426) and Fraass breaking point (EN 12593). Tests at 25% (w/w) content of PMMA/ATH composite dust in the bitumen showed a 7.4 ° C increase in temperature softening point (EN 1427) and a 4 ° C decrease in temperature of Fraass breaking point (EN 12593) with respect to the base bitumen. Rheological tests at 60 ° C showed an increase in rut formation parameter of G*/sin (). Basic bitumen has rut formation parameter 1330 Pa, at 25% (w/w) content of PMMA/ATH composite dust in the bitumen measured value of rut formation parameter 3650 Pa, which is almost three times greater resistance to the formation of ruts as base bitumen.
F.33 Slovenian patent
COBISS.SI-ID: 5121818The target of the presented research is Cathepsin K (Cat K), a lysosomal cysteine protease that plays an important role in many severe diseases, which makes inhibition of Cat K a potentially attractive therapeutic approach. Several compounds, including balicatib, passed preclinical studies and were tested in clinical trials as perspective Cat K inhibitors. Balicatib proved as a potent inhibitor, however, the side effects caused by the accumulation of drug candidate in the lysosomes of human skin fibroblast as a consequence of its basic nature prevented further drug development. We have introduced a combination of chemometrics and virtual combinatorial library approaches employed in order to better assess the correlation between molecular structure and biological activity (Cat K inhibition) of benzamide-containing aminonitriles and predict potential new inhibitors with less side effects connected to basic nature of some clinically trailed drug candidates. A QSAR model was constructed based on the Counter-propagation artificial neural network (CP-ANN), which was trained with a set of diversely substituted aminonitriles of known experimentally determined activities. Interpretation of selected molecular descriptors was performed by evaluation of overlapping contours of the Kohonen levels with the CP-ANN response surface and with clusters of objects in the Kohonen top-map (SOM). The developed QSAR model is proposed to be used to design new non-toxic Cat K inhibitors. The synthesis of predicted most potent inhibitors is in progress and will be subjected to additional testing.
B.04 Guest lecture
COBISS.SI-ID: 36971781QSAR methods are gaining greater acceptance as an alternative to animal testing, to fill data gaps for hazard communications (U.N. Globally Harmonized System for chemical labeling) and data submissions for regulatory purposes (European chemical regulation REACH). QSAR software models are constructed using data systems involving molecular descriptors, specific to toxic endpoints, and reference databases comparing known data references, with unknown data points. VEGA, TEST, and OECD QSAR Toolbox are three widely known publicly available software platforms. Most of the applied computer models have been developed under consideration of OECD principles for validation of (Q)SAR models used for regulatory purposes. In the presentation an accent is placed on VEGA platform and the five toxicological endpoints predicted by CAESAR programs (bioaccumulation factor, mutagenicity, carcinogenicity, developmental toxicity and skin sensitization). Three data sets are discussed: the set of poly aromatic hydrocarbon (PAH), the set of (con)azoles and the set of compounds taken from the CosIng inventory. Variations in software performance, including specificity, sensitivity, and strength of prediction (regression) are evaluated. For further analysis of applicability domains and similar compounds the chemometrical methods for clustering are applied, like principal component analysis and Kohonen networks.
B.04 Guest lecture
COBISS.SI-ID: 5364250Lecture at the Workshop on Cellular Membrane Transport, Università degli studi di Trieste, 3rd December, 2013. Trieste: Università degli studi di Trieste, 2013.
B.05 Guest lecturer at an institute/university
COBISS.SI-ID: 5419546