Novel computational approaches for pathogen detection

Computational methods are an essential pillar in deriving information from biological data which can inform pathogen surveillance and diagnostics. Metagenomics has great potential to advance our understanding of bacteria and for pathogen surveillance. However, high resolution taxonomic classification and the assignment of genes to bacterial strains still present major hurdles. For viruses, surveillance can be aided by group testing, which can greatly reduce testing resources required. Yet, no clear pathway from method development to clinical application exists and there is still much potential for theoretical developments improving the information gained from group testing. This thesis addresses the above points in two parts. First, following a benchmarking study of available tools, a novel metagenomic strain identification tool is developed. Results show state-of-the-art performance on simulated and mock metagenomic data. Additionally, accurate assignment of accessory genes to strains in metagenomes using the tool is possible. Secondly, a new group testing decoder is developed and evaluated within a novel validation framework for high-throughput group testing validation. This study shows the in vitro feasibility of group testing using a non-binary decoder. The presented decoder represents a first step towards designing a group testing decoder for multiplexed group testing, and future research avenues are discussed in detail. Finally, an exact expression for estimating prevalence from array testing data is derived. For the same number of tests, the array testing estimator proves to be more accurate than individual testing. Altogether, this thesis contributes valuable new methods for the analysis of metagenomics and group testing data. The metagenomics approach has the potential to provide new, interpretable insights into bacterial strains in microbiomes and the genes they carry. The work on group testing and prevalence estimation shows their usefulness for surveillance testing and highlights the potential of widespread implementation of group testing in public health applications.