Phylogenomics and gene association analysis support probiotic potential in a subset ofisolates.

Abstract

Mastitis is a major disease in dairy cattle, often caused byinfection and increasingly complicated by antimicrobial resistance. As a result, bacterial strains able to effectively provide colonization resistance in the bovine mammary gland are being explored as potential probiotics to reduce reliance on antimicrobial treatments. Among these, some non-.staphylococci species have shown promise; however, their effects appear to be highly strain-dependent. In this study, a combination of machine learning approaches alongside standard genomic and phylogenetic analyses was used to compare and select putative probioticstrains. The genomes of 82&#x2009;S.and 16 closely relatedisolates from milk samples obtained from the Canadian Bovine Mastitis Research Network were assembled and annotated. In addition to identifying orthologous gene families and reconstructing phylogenetic relationships, each genome was screened for virulence factors, antimicrobial resistance (AMR) genes and bacteriocins. Random forest modelling and association rule learning were then applied to identify combinations of genes associated with isolates from milk samples collected from quarters exhibiting low inflammation, assessed using somatic cell counts (SCC). This approach identified 63 genes that frequently co-occurred in isolates from low SCC samples (low SCC <25,000&#x2009;cells&#x2009;ml) but were largely absent in those from high SCC samples (&#x2265;200,000&#x2009;cells&#x2009;ml). These gene sets were used as biomarkers in conjunction with phylogenetic and clustering analyses to guide the selection of a subset ofisolates with potential probiotic properties.