Protein Local Structure Prediction Using Support Vector Machine

KWONG, PEI YEEN (2010) Protein Local Structure Prediction Using Support Vector Machine. Other thesis, Universiti Teknologi Malaysia.


Download (1965Kb) | Preview


Structural bioinformatics is getting essential these days. Secondary structure segments are the building blocks of protein local structure. Protein local structure prediction leads to a better three dimensional protein structure and function prediction. Predicting protein local structure is capable to predict protein global structure at ease based on amino acid sequence. However, most related works are conducted using fixed segment length which possibly leads to bias results. Therefore this experiment is conducted to determine the optimal segment length. In this research, residue score is quantified using the hybrid of conservation score and propensity score named CP. Each amino acid is assigned a secondary structure alphabet using DSSP and STRIDE. The residue scores with their respective structural class are sliced into segments using sliding window technique and then modified to feature vector and feature class. Next, Support Vector Machine (SVM) performs classification based on feature vector and feature class to predict protein local structure. This experiment utilizes RS126 dataset (Rost and Sander, 1994) and two fold cross validation. The optimal length of protein local structure for helix is 9, strand is 19 and coil is 5 consecutive residues based on the performance of accuracy and proportion of support vectors. The optimal Secondary Structure Assignment Method (SSAM) that gives the best performance in accuracy, specificity and sensitivity is determined.

Item Type: Thesis (Other)
Subjects: Q Science > QA Mathematics > QA75 Electronic computers. Computer science
Divisions: Faculty of Computer Science and Information System > Bioinfomatic
Depositing User: Unnamed user with email
Date Deposited: 04 Jul 2013 06:59
Last Modified: 04 Jul 2013 06:59

Actions (login required)

View Item View Item