Text

theses

Sequence alignment techniques have been developed into extremely powerful tools for identifying the folding families of the new protein sequences that are becoming available through the genome projects. It is known that proteins sharing more than 30 sequence identity over the majority of their length have a high probability of adopting the same fold and conventional sequence comparison methods easily detect these similarities. Yet, analysis of the relationship between sequences and structure similarity have shown that around the "twilight zone" or sequence identity of about 17 to 25%, the relationship between primary and tertiary structure becomes problematic. Two proteins may have identical topologies (folds) without sharing detectable sequence similarities. Such fold similarities will normally not be found until both protein 3-D structures have been determined experimentally. Alternative methods which bypass the high cost of experimental structure elucidation include homology modeling and fold recognition techniques which are based on observation that during evolution, folds vary much less than amino acid sequences. Furthermore, the organization of secondary structure elements in space determines the three-dimensional fold of a protein. Secondary structure information therefore may be used to identify folding families when the amino acid sequence identities within the folding families are low. Yet, this information about protein structure topology derived only from the sequence of secondary structure states of residues has not yet been fully exploited. The goal of this work is to show that protein topology can indeed be recognized from the sequence of secondary structures which subsequently can be used as a criterion for fold homology. We have constructed a secondary structure similarity matrix based on a database of three dimensionally aligned proteins (3D_ali). Alignments were then carried out with different mixes of amino acid and secondary structure information to ascertain the reliability of the methodology. We used the SCOP40 database, where only the PDB sequences that have 40% sequence of less are included to evaluate homology detection by the combined amino acid and secondary structure alignments. The results presented in this thesis show that the sequential arrangement of the secondary structure contains significant fold information in addition to the primary structure alone. The quality of the secondary structure information (true VS predicted) has a large influence on the results. Incorporating predicted secondary structure information for six small genomes yields enhancement of homology detection by 20% at a low error rate. Using alignments carried out with secondary structure information, we also clustered proteins with low amino acid sequence identity according to their secondary structures. To ascertain the reliability of this clustering method, we compared it to the other already existing databases such as SCOP, CATH and 3D_ali using statistical analysis which tested the coverage, specificity and error-per-query for each of the databases used in this study. Finally, we also ran a 200 picosecond molecular dynamics simulation on a small 50 amino acid peptide, human TGF-alpha (hTGF-alpha). The generalized order parameters were then calculated from the MD simulation trajectory for both the N-H and C-alpha bonds. During the MD simulation, we observed systematic difference between the order parameters for the N-H bonds and those for C-alpha bonds. Consistent with the results of the experimental studies, the difference was found to be highly correlated with the crankshaft librations of the peptide planes.

ETD_4936

doi:10.7282/T3513W71

Ph.D.

Includes bibliographical references

by Addi R. Fadel

ETD doctoral

The author owns the copyright to this work.