Whole shotgun metagenomics sequencing allows researchers to retrieve information about all organisms in a complex sample. This method enables microbiologists to detect pathogens in clinical samples, study the microbial diversity in various environments, and detect abundance differences of certain microbes under different living conditions. The emergence of nanopore sequencing has offered many new possibilities for clinical and environmental microbiologists. In particular, the portability of the small nanopore sequencing devices and the ability to selectively sequence only DNA from interesting organisms are expected to make a significant contribution to the field. However, both options require memory-efficient methods that perform real-time data analysis on commodity hardware like usual laptops. In this thesis, I present new methods for real-time analysis of nanopore sequencing data in a metagenomic context. These methods are based on optimized algorithmic approaches querying the sequenced data against large sets of reference sequences. The main goal of those contributions is to improve the sequencing and analysis of underrepresented organisms in complex metagenomic samples and enable this analysis in low-resource settings in the field. First, I introduce ReadBouncer as a new tool for nanopore adaptive sampling, which can reject uninteresting DNA molecules during the sequencing process. ReadBouncer improves read classification compared to other adaptive sampling tools and has fewer memory requirements. These improvements enable a higher enrichment of underrepresented sequences while performing adaptive sampling in the field. I further show that, besides host sequence removal and enrichment of low-abundant microbes, adaptive sampling can enrich underrepresented plasmid sequences in bacterial samples. These plasmids play a crucial role in the dissemination of antibiotic resistance genes. However, their characterization requires expensive and time-consuming lab protocols. I describe how adaptive sampling can be used as a cheap method for the enrichment of plasmids, which can make a significant contribution to the point-of-care sequencing of bacterial pathogens. Finally, I introduce a novel memory- and space-efficient algorithm for real-time taxonomic profiling of nanopore reads that was implemented in Taxor. It improves the taxonomic classification of nanopore reads compared to other taxonomic profiling tools and tremendously reduces the memory footprint. The resulting database index for thousands of microbial species is small enough to fit into the memory of a small laptop, enabling real-time metagenomics analysis of nanopore sequencing data with large reference databases in the field.
