Interpretable Model for Antibiotic Resistance Prediction in Bacteria using Deep Learning
Manoj Jha1, Akshay Kumar Kawale2 and Chandan Kumar Verma11Department of Department of Mathematics, Bioinformatics, Computer Applications MANIT, Bhopal, India.
2Department of Bioinformatics, MANIT, Bhopal, India.
Abstract: The identification of Antibiotic resistance in bacteria is a key step of improvement in the field of drug discovery and vaccinology. We present a method for this task that relies on a k-mer representation of genomes and a deep learning algorithm that produces interpretable models. The method is computationally accessible and well-suited for whole genome sequencing studies. Deep learning is an application of machine learning that uses a cascade of many layers of nonlinear processing units for extracting features and transforming it. The existing approaches for predicting antibiotic resistance genes in bacteria is not efficient enough whereas machine learning proves to be more effective than traditional methods. Our study relies on a k-mer representation method. In computational genomics, k-mer refers to all the possible subsequence (of length k) from a read obtained through DNA sequencing. The study generates the result with the help of features like coverage and depth that tells us about resistivity of the bacteria against the antibiotic. The accuracy of the model varies from 93% to 97%. The method was validated by generating models that predicted the antibiotic resistance of bacteria. The model is accurate, faithful to biological pathways targeted by the antibiotics, and they provide insight into the process of resistance acquisition. The model is computationally scalable and well suited for whole genome sequencing studies.
Keywords: Antibiotic Resistance; Bacteria;Deep Learning; Genomics Machine learning; Back to TOC