Open X-Embodiment: Robotic learning datasets and RT-X models

Moraxella bovoculi is the bacterium most often cultured from ocular lesions of cattle with infectious bovine keratoconjunctivitis, also known as bovine pinkeye. Some strains of M. bovoculi contain operons encoding for a repeats-in-toxin (RTX) toxin, which is a known virulence factor of multiple veterinary pathogens. We explored the utility of MALDI-TOF MS and biomarker detection models to classify the presence or absence of an RTX phenotype in M. bovoculi. Ninety strains that had undergone whole genome sequencing were classified by the presence or absence of complete RTX operons and confirmed with a visual assessment of hemolysis on blood agar. Strains were grown on Tryptic Soy Agar (TSA) with 5% sheep blood, TSA with 5% bovine blood that was supplemented with 10% fetal bovine serum, 10 mmol/LCaCl2, or both. The formulations were designed to determine the influence of growth media on toxin production or activity, as calcium ions are required for toxin secretion and activity. Mass spectra were obtained for strains grown on each agar formulation and biomarker models were developed using ClinProTools 3.0 software. The most accurate model was developed using spectra from strains grown on TSA with 5% bovine blood and supplemented with CaCl2, which had a sensitivity and specificity of 93.3% and 73.3%, respectively, regarding RTX phenotype classification. The same biomarker model algorithm developed from strains grown on TSA with 5% sheep blood had a substantially lower sensitivity and specificity of 68.0% and 52.0%, respectively. Our results indicate that MALDI-TOF MS biomarker models can accurately classify strains of M. bovoculi regarding the presence or absence of RTX toxin operons and that agar media modifications improve the accuracy of these models.