Overview Of Artificial Neural Networks

Good Essays

Introduction Artificial neural networks are a class of computational structures (Lesk, 2013) made up of several processing elements, called artificial neurons that are connected and organized in layers (Larder et al., 2007). They are capable of generating models for the detection of non linear functions(..). Their algorithms are extensively applied in biology and medicine to solve complex problems, more specifically for prediction or classification of solutions or to refine methodological aspects. (Florence and Balasubramanie, 2010). Human immunodeficiency virus (HIV) is a retrovirus that can lead to acquired immunodeficiency syndrome (AIDS). (Kim et al., 2010). It is a disease in which the body immune system weakens progressively, …show more content…

Artificial neural networks have been used extensively as a complimentary bioinformatics tool to make approximations of the cleavage site activity and specificity. First uses of ANNs to solve the problem The aim of first research study was to develop a classification model that, given a sequence of eight amino acids, could discriminate between sequences which are either cleavable or uncleavable by the HIV- 1 protease. (Kim et al., 2010). The neural learning algorithms used most frequently was back-propagation neural networks (BPNNs) (Thomson et al., 2003) because it performs well on prediction problems. (Sibanda and Pretorius, 2012). When BBNN was used for the prediction of the HIV-1 protease cleavage site, it gave a prediction accuracy 92%(Thomson et al., 2010) However, one of the major disadvantages of using ANNs to analyse biological data referred to the impossibility of most ANNs of recognizing non- numerical features like amino acids. Hence an encoding process to model the amino acids was preferable. (Thomson et al., 2003) The advantage of the Bio basic functional neural networks The peculiarity of this algorithm relates to its ability to recognise amino acids directly. Thus, avoiding the use of 20 binary bits to represent each amino acid is advantageous. (2003) The prediction accuracy of BBFNN was proved in a research study using 362 HIV protease sequences, where 114 were with cleavage sites and

Get Access
Get Access