基于遗传算法的子网络生物标记优化方法

It has shown remarkable success in the prediction of breast cancer metastases based on the combination of gene expression profiles and protein-protein interaction networks. As an important step of network-based methods, the problem of effectively identifying predictive subnetwork markers remains a great challenge. Existing methods typically use greedy search algorithms, which may not find out the optimal subnetwork markers and accordingly impair the performance of the learning machines. In this paper, we propose a genetic algorithm to improve the subnetwork markers identified by an existing greedy search method. We demonstrate that the discriminative power of the optimized subnetwork markers are significantly higher than the original subnetwork markers, and we can get higher classification performance of the learning machines when using the optimized subnetworks as predictive features via 3 popular machine learning approaches (logistic regression, support vector machine and random forest). According to the comparison of different classification approaches, random forest algorithm with the optimized subnetwork markers shows the highest classification results among all the methods compared, as well as good reproducibility of the optimized sub-networks for better identifying the unknown breast cancer metastases data.