Modelling protein prediction using data mining techniques

Currently, many machine learning methods are employed in protein secondary structure prediction that includes neural network, support vector machine, bayesian segmentation, genetic algorithm and others. However, researchers have difficulty to determine and identify the right machine learning methods to be used because of the complex structure of protein which are primary, secondary, tertiary and quaternary structure. This is due to the different characteristics of protein which are structure, function, charge, acidity, hydrophilicity and molecular weight and the different type of shape of the protein which are alpha helices, beta strands and coils. Hence, researcher would like to propose a model that attempts to assist other researchers in order to choose a suitable classification technique to be used for protein structure prediction. The analysis is derived by excavating literature on protein secondary structure prediction starting from the year 1987 until 2006, focused on supervised learning algorithm. The model demonstrates general flow for protein secondary structure from sequence to structure (Q2T), structure to structure (T2T) and reliability index. As a result, a model of data mining techniques used in protein secondary structure prediction is built. Therefore, future work can be tackled by using more past work that had been applied in data mining techniques, other than protein structure but in other tasks like function prediction, location prediction, protein interaction and protein annotation, in order to get a better view of it.