Consensus Prediction of Protein Secondary Structures

I dedicate this thesis to my parents. They raised me up, provide me happiness, cultivate my personality, and provide me the best education. I dedicate this thesis to my grandmother, a diligent and kind Chinese woman, who made great contributions to the whole family, and left us in 2004 when I was pursuing this degree in Canada. I wish her peace forever. ii Protein structure prediction is one of the most significant problems in bioinformat-ics. Currently, there are some tools which can predict protein secondary structure, or find protein structural motifs and some specific structure segments. However, their results are sometimes different or contradictory. CISPred is a consensus protein structure prediction system which integrates results in order to provide overall consensus predictions of protein secondary struc-tures. The average accuracy of CISPred predictions is 82.6 % on a dataset con-taining 109 CASP sequences, and 89.3 % on a dataset containing 1758 sequences. iii Acknowledgements I sincerely appreciate my supervisors, Dr. Patricia Evans and Dr. Virendra Bhavsar. They impart their knowledge, direct my research, and provide finan-cial support. Dr. Virendra Bhavsar and Dr. Patricia Evans are professors with profound knowledge and experience, and have been respectful mentors. The two years of study and research with them have been one of the best periods in my life.