Gene Mutation Classification Using CNN and BiGRU Network

Designing medical solutions based on genomic information can help achieve precision medicine. Currently, clinical pathologists have to manually review and classify each gene mutation according to the evidence in the clinical literature, which is a time-consuming work. In this paper, we propose a novel machine learning algorithm that can classify gene mutations based on clinical evidence (text). The algorithm model combines convolutional neural network (CNN) with Bidirectional Gated Recurrent Unit (BiGRU) sequence neural network, in which CNN is used as feature extractor of semantic information and BiGRU is used as mapping device of context semantic structure. This model can analyze text from two aspects-semantic information features and context semantic structure. We train and evaluate the model using a real-world dataset. Experimental results show that our model outperforms traditional single neural network model and some mixed neural network models.