ABSTRACT
We assessed the diagnostic yield of metagenomics urine sample testing in patients with urological symptoms. We conducted microbiome analysis of 86 female urine samples that included 17 healthy controls and 69 patients. Natural language processing (NLP), a subfield of artificial intelligence, was used to create a pathogen identification tool, Xplore-AI, to assess the potential pathogens in all of the samples. Meanwhile, report summaries that were written by infectious disease experts were compared to the NLP results to investigate its accuracy. The results showed that the NLP system reported 97% of patient samples had at least one pathogen over three standard deviations from values found in in healthy controls. Similarly, 84% of patients had two or more classified pathogens. These diagnostic percentages were consistent with the infectious disease expert summaries. However, some pathogens like Aerococcus urinae were present in 13 patient samples, but only reported in one summary. In conclusion, this study demonstrated the high diagnostic yield in females with urological symptoms following metagenomic analysis and the ability of using an NLP-based system to identify pathogens to improve the accuracy of the reportable species.
