Issue No. 01 - January/February (1999 vol. 14)
DOI Bookmark: http://doi.ieeecomputersociety.org/10.1109/5254.747907
<p>In the 45 years since the discovery of DNA'S helical structure, scientists have made great strides in exploring human DNA's structure and locating human genes. As part of that effort, advanced recombinant DNA and gene-mapping techniques developed over the last two decades have led to an unprecedented effort to map and sequence the entire human genome under the auspices of the Human Genome Project.</p> <p>The huge amount of data the HGP produces will require high-performance computing and more intelligent computer algorithms for analysis and inference, needs that the emerging field of computational molecular biology is addressing. Interest has been growing in exploring such AI tools as search algorithms, machine-learning techniques, and knowledge-based systems for DNA sequence analysis. Recently, the neural-network model has emerged as a promising AI technique in this regard because this approach might well embody important aspects of intelligence not captured by symbolic and statistical methods.</p> <p>An important direction in this work involves integrating multiple, fundamentally different AI approaches into single hybrid intelligent systems, which let each component perform the tasks for which it is best suited. Integrating symbolic knowledge into a neural network to create a knowledge-based neural network has quickly become an important hybrid-intelligence research area. Empirical observations indicate that such systems can outperform both neural-network and symbolic approaches. This integrated approach involves mapping a set of symbolic rules that encode available domain knowledge into the neural computing architecture and requires innovative methods for extracting symbolic knowledge from a trained neural network.</p> <p>Called expert networks in some cases, these knowledge-based neural networks perform as well as human experts and often exhibit characteristics of a traditional symbolic expert system. In one implementation, the expert network combines a neural network's computational framework with an expert system's inference engine. We call a neural network that bases the activation function on the certainty factor model of Mycin-like systems a CF-net. This article describes my successful attempt to apply the CF-net to DNA sequence analysis. As this article shows, the CF-net uses as the system input the DNA nucleotide sequence rather than predefined relative frequency measures as in existing exon-prediction systems.</p>
expert network, neural network, certainty factor, DNA analysis, spline junctions, gene identification.
LiMin Fu, "An Expert Network for DNA Sequence Analysis", IEEE Intelligent Systems, vol. 14, no. , pp. 65-71, January/February 1999, doi:10.1109/5254.747907