Pathogenic virus detection method based on multi-model fusion

Abstract

Identifying viruses is crucial for pandemics. Detecting pathogenic viruses is difficult and usually needs to spend lots of time. Here, we propose a multi-model fusion method for viruses pathogenic detection(MMFPV). We use CatBoost, Reverse-complement Convolutional Neural Networks, and, Reverse-complement Long Short-Term Memory(RC-LSTM) as base models, then, using stacking to combine these models. This method directly detects pathogenic viruses from next-generation sequencing data without virus databases. The result shows that this novel detection method has good capabilities.