A Recurrent Neural Network approach for whole genome bacteria classification

Abstract

The classification of bacteria plays an essential role in multiple areas of research. Those areas include experimental biology, food and water industries, pathology, microbiology, and evolutionary studies. Although there exist methodologies for classification - such as mass spectrometry, single-nucleotide polymorphisms, microscopic morphology, and neural network approaches - a transition to a whole genome sequence based taxonomy is already undergoing. Next Generation Sequencing helps the transition by producing DNA sequence data efficiently. However, the rate of DNA sequence data generation and the high dimensionality of such data need faster computer methodologies. Machine learning, an area of artificial intelligence, has the ability to analyze high dimensional data in a systematic, fast, and efficient way. Therefore, we propose a sequential deep learning model for bacteria classification. The proposed neural network exploits the vast amounts of information generated by Next Generation Sequencing, in order to extract a classification model for whole genome bacteria sequences. A distributed representation based on k-mers of k={3,4,5} provided an efficient encoding for the bacterial sequences. The classification model relies on a bidirectional recurrent neural network architecture. It generates an accuracy of 0.99455 +/- 0.00281 for 14 species, 0.95031 +/- 0.00469 for 48 species, and 0.89107 +/- 0.00392 for 111 species. After validating the classification model, the bidirectional recurrent neural network outperformed other classification approaches, such as Naive Bayes and Feedforward neural network. The proposed model provides an automated identification method. It infers species for bacterial whole genome sequences and it does not require any manual feature extraction.