A Genome-based Machine Learning Model for Safety Assessment of Microorganisms

The identification of probiotic and pathogenic microorganisms is crucial across various fields, including food safety, environmental assessment, and medical diagnostics. However, current technologies often fall short in terms of efficiency and convenience. Here, we develop iMicrobes, a genome-based machine learning model designed to achieve rapid and accurate microbial safety identification. By leveraging both nucleotide and codon features from genomic sequences, iMicrobes feeds these data into three separate Support Vector Machine (SVM) frameworks, achieving an accuracy of over 0.98, requiring only 83 features for neutral microbes, 45 for probiotics, and 135 for pathogens, respectively. Utilizing iMicrobes, we have identified a total of 55,412 probiotic and 1,088,863 pathogenic genomes within genomic databases.