show Abstracthide AbstractN6-methyladenosine (m6A) is among the most common bacterial base modifications. Detection at single base resolution is possible using single-molecule sequencing with PacBio and more recently, nanopore sequencing. However, accuracy is variable and hasn't been well characterized across platforms. Here, we compare sites detected using PacBio data and SMRT Tools to those detected using nanopore data across eight species in a microbial reference community. We also use Illumina data from individual strains and the full community to generate and help validate assemblies for the reference community. With m6A sites validated using PacBio, we train and test a new neural network-based method to detect m6A from nanopore data and show it can improve detection depending on sequence context.