**Background:** Probiotics are increasingly used in livestock to improve health and growth performance and reduce antibiotic use. Accurate labelling of probiotic products is critical for safety and efficacy, yet studies have reported high rates of mislabelling (40–47% in one European survey). Conventional quality control methods (culture, PCR, short-read sequencing) have limitations in resolution and accuracy. Long-read sequencing technologies, such as Oxford Nanopore, offer potential for improved microbial identification in mixed communities. This study aimed to evaluate two long-read-only workflows (full-length 16S rRNA gene sequencing and metagenomic sequencing) for assessing the microbial composition, safety, and quality of commercial animal probiotic feed supplements available in Thailand.
**Methods:** Two commercial animal probiotic products were collected in Thailand in January 2022: Product A (liquid, labelled with six bacterial strains including Lactobacillus acidophilus, Lactobacillus fermentum, Lactobacillus paracasei, Lactobacillus plantarum, Bacillus licheniformis, and Bacillus subtilis) and Product B (powder, labelled with Bacillus coagulans, Bacillus subtilis, and Saccharomyces cerevisiae). Metagenomic DNA was extracted using the ZymoBIOMICS DNA miniprep Kit. Full-length 16S rRNA genes were amplified using the 16S Barcoding Kit (27F/1492R primers), and metagenomic libraries were prepared using the Rapid Barcoding Sequencing Kit. Sequencing was performed on a MinION (Mk1C) with R9.4.1 flow cells. Base-calling used Guppy v6.0.1 in super accuracy mode. Taxonomic classification of 16S data was performed using NanoCLUST (against NCBI RefSeq) and Kraken2 v2.1.2 (against RefSeq PlusPFP-8). Metagenomic data were classified using Kraken2 and also assembled into metagenome-assembled genomes (MAGs) using metaFlye v2.9, polished with Racon and medaka, and binned with MetaBAT2. MAG taxonomy was assigned using GTDB-Tk v1.5.1 and NCBI BLAST. AMR genes were annotated using ABRicate against NCBI AMRFinderPlus, ResFinder, and VFDB. BGCs and bacteriocins were identified using antiSMASH v6.0 and BAGEL4.
**Key Results:** Approximately 210 Mbp and 1.1 Gbp (Product A) and 253 Mbp and 469 Mbp (Product B) of 16S amplicon and metagenomic sequence data were generated, respectively. For Product A, all tools identified L. paracasei as the dominant taxon, followed by L. plantarum. Critically, L. acidophilus—listed on the label—was not detected by any method. B. licheniformis was recovered only by 16S-Kraken2 at 0.1% relative abundance. For Product B, 16S amplicon sequencing identified Bacillus subtilis and B. coagulans, but the yeast Saccharomyces cerevisiae was not detected by 16S; metagenomic data (MAGs) enabled its identification. Unlisted species were identified in both products: Bacillus velezensis comprised up to 18% (Product A) and 40% (Product B) depending on the tool used. Other unlisted species included B. licheniformis and Caldibacillus thermoamylovorans in Product B. MAGs recovered seven genomes from Product A and four from Product B, with completeness ranging from 17% to 99.4% (average 80.7%) for Product A and 3.4% to 99% (average 67.4%) for Product B, and contamination <3.6% and <0.2%, respectively. ANI values ranged from 97–99%. AMR gene analysis revealed two AMR genes in Product A and nine in Product B. The chloramphenicol resistance gene cfr(B) was found in both products. Product B additionally harboured genes for aminoglycosides [ant(4')-Ib, aadK], chloramphenicol (cat4), macrolides [erm(34), erm(D), mphK], and quinolone (qnrD1). Product A contained a tetracycline efflux gene tet(L). No virulence factors were identified. BGC analysis identified seven putative BGCs in Product A and ten in Product B, including NRPS, RiPPs, and transAT-PKS clusters. Bacteriocin analysis revealed eight and eleven classes from Products A and B, respectively.
**Clinical Implications:** This study demonstrates that long-read-only Nanopore sequencing workflows can effectively assess the microbial composition and safety of commercial animal probiotic products. The detection of labelling discrepancies (missing declared strains and presence of unlisted bacteria) and AMR genes highlights the need for improved quality control in the probiotic industry. The presence of B. velezensis as a dominant unlisted taxon in both products suggests potential misidentification due to high genomic similarity within the B. subtilis group complex and the use of outdated databases. The identification of bacteriocin and BGC genes indicates potential beneficial properties that could be leveraged as alternatives to antibiotic growth promoters. The authors recommend that regulatory bodies and producers adopt long-read sequencing approaches for routine quality assurance to ensure product safety, accuracy of labelling, and consumer trust.