Treffer: Comparison of neural network optimization approaches for studies of human genetics

A major goal of human genetics is the identification of susceptibility genes associated with common, complex diseases. The preponderance of gene-gene and gene-environment interactions comprising the genetic architecture of common diseases presents a difficult challenge. To address this, novel computational approaches have been applied to studies of human disease. These novel approaches seek to capture the complexity inherent in common diseases. Previously, we developed a genetic programming neural network (GPNN) to optimize network architecture for the detection of disease susceptibility genes in association studies. While GPNN was a successful endeavor, we wanted to address the limitations in its flexibility and ease of development. To this end, we developed a grammatical evolution neural network (GENN) approach that accounts for the drawbacks of GPNN. In this study we show that this new method has high power to detect gene-gene interactions in simulated data. We also compare the performance of GENN to GPNN, a traditional back-propagation neural network (BPNN) and a random search algorithm. GENN outperforms both BPNN and the random search, and performs at least as well as GPNN. This study demonstrates the utility of using GE to evolve NN in studies of complex human disease.