This thesis explores aspects protein function, structure and sequence by computational approaches. A comparative study of definitions of protein secondary structures was performed. Disagreements in assignment resulting from three different algorithms were observed. The causes of inaccuracies in structure assignments were discussed and possibilities of projecting protein secondary structures by different structural descriptors were tested. The investigation of inconsistent assignments of protein secondary structure led to a study of a more specific issue concerning protein structure/function relationships, namely cis/trans isomerisation of a peptide bond. Surveys were carried out at the level of protein molecules to detect the occurrences of the cis peptide bond, and at the level of protein domains to explore the possible biological implications of the occurrences of the structural motif. Research was then focussed on andalpha;-helical integral membrane proteins. A detailed analysis of sequences and putative transmembrane helical structures was conducted on the ABC transporters from different organisms. Interesting relationships between protein sequences, putative a-helical structures and transporter functions were identified. Applications of molecular dynamics simulations to the transmembrane helices of a specific human ABC transporter, cystic flbrosis transmembrane conductance regulator (CFTR), explored some of these relationships at the atomic resolution. Functional and structural implications of individual residues within membrane-spanning helices were revealed by these simulations studies.

The automated protein structure analysis (APSA) has been developed that describes protein structure via its backbone in a novel way. APSA generates a smooth line for the backbone which is completely described using curvature κ and torsion τ as a function of arc lengths. Diagrams of κ(s) and τ(s) reveal conformational features as typical patterns. In this way ideal and natural helices (α, 310 and π) and β-strands (left and right-handed, parallel and antiparallel) can be rapidly distinguished, their distortions classified, and a detailed picture of secondary structure developed. Such foundations make it possible to qualitatively and quantitatively compare domain structure utilizing calculated κ(s) and τ(s) patterns of proteins. Focusing on the torsion diagrams alone, 16 regions of τ(s) values that correspond to unique groups of conformations have been identified and encoded into 16 letters. The entire protein backbone is described, effectively projecting its three-dimensional (3D) conformation into a one-dimensional (1D) string of letters called the primary code (3D-ID projection), which is APSA's conformational equivalent of a protein's primary structure. The secondary structure is obtained from specific patterns of the primary code (resulting in secondary code). The letter code is used to describe supersecondary structure, which involves a unique characterization of the tum. It contains sufficient information to reconstruct the overall shape of a protein in an unambiguous 1D→3D translation step. Therefore, it is possible to classify supersecondary structure with the help of the letter code in form of a novel labeling system (F.#.M.X.O.L.N.R.U.S) that collects information on the relative orientations of tum and flanking structures (helices, strands). The overall shape of supersecondary structure is obtained by partitioning the surrounding space into octants and cones and assigning the parts of a supersecondary structure to these sub spaces via its labels. This approach can be easily extended to tertiary structure.