DEMINERS enables multiplexing, precise basecalling, and isoform regulation in nanopore direct RNA sequencing

Nanopore direct RNA sequencing (DRS) advances RNA biology but is limited by high input and costs, relatively low basecalling accuracy, and lack of multiplexing capabilities. Here we developed DEMINERS, which combines a de-multiplexing workflow using a Random Forest classifier, an optimized basecaller using a convolutional neural network (CNN), and comprehensive downstream analysis for RNA isoform profiling at single-read level. It also integrates variant calling, genome assembly and RNA modification. When applying to metagenomic analysis of 24 COVID-19 patients and comparative studies of two gliomas types and different malaria parasites stages, we revealed key distinctions in nasopharyngeal and oropharyngeal microbiomes and potential regulation of m6A modifications in alternative splicing in cancer and parasites. Our method enhances the scalability and accuracy of DRS and demonstrates its potential in de novo assembly of virus genome and in uncovering rich multi-omics data of different samples from a single DRS run.