Embedded System for Exon Predictor at DNA Sequences Using Hidden Markov Model

• Main Authors: H. H. Arifin, S.S. Samsudin
• Format: Conference Paper
• Language: English
• Published: 2013
• Subjects: Exon predictor, DNA sequences, Hidden Markov Model.
• Online Access: http://ddms.usim.edu.my/handle/123456789/6162

Artificial Intelligence has been tried to understand by philosophers for a thousand years. Many researches proved its truth in intelligence machine or computational programs. The complexity of Al created a new family in Bioinformatic and ioscience. Its elements are implemented by basic algorithms such as combination of repetition, condition, decision making and. nonhuman basic logic recursive. Their members are consisted of Neural Networks. Genetic Algorithms, Hidden Markov Model, and Stochastic ContextFree Grammars. HMM is a powerful statistical tool to predict an event actually, such as Deoxyribonucleic Acid whereas formed by 4 kinds of bases Adenine (A), Thymine (T), Cytosine (C) and Guanine (G). The growing up of embedded system motived HMM to be embedded in hardware base. Because of that, it is possible to inject the smart component into a stand alone with PlatformBased Design board. Moreover, the attending of Open Source Software also effected to this development area. For the smart component, the DNA sequences are separated into several states such as 5, 7 and 9 states whereas the numbers of states can be increased randomly. Then, it is essentials to generates the emission matrix to shows probabilities of base distribution depend on the numbers of state and the transition matrix which predicted by human. The combination of numbers of state, emission and transition matrix formed a HMM's model. The HMM training and testing used the ForwardBackward and Viterbi algorithms. The training results for 7 states model and 152 samples found the local maximum likelihood is 1.585385 which showed the suitability between model and observation sequences, beside the estimation of emission and transition matrix. Then, the testing section presented maximum Correlative Coefficient of model is 0.051351. But, we have to calculate 4 required parameters before that: True positive (TP), True Negative (TN), False Positive (FP) and False Negative (FN).