Applications of next generation sequencing to microbial evolution

Next generation sequencing (NGS) is defined by parallel sequencing of short DNA fragments, or ‘reads’. In ‘deep sequencing’ studies, aligned NGS reads are treated as a population sample, with genetic variation in individual reads being meaningful. Deep sequencing has revolutionised the study of microbial evolution. However, distinguishing single nucleotide variants (SNVs) from sequencing errors remains challenging.In this thesis, I first show the power of deep sequencing in a hepatitis C virus (HCV) study. Using published methods and manual analysis, SNVs with frequencies down to 0.1% were detected. Phylogenetic analysis indicated early HCV evolution is driven by two bottlenecks, associated with transmission and immune response respectively.The HCV study's SNV calling methods are then developed and improved, by introducing a statistical test of strand bias (to remove systematic errors) combined with probabilistic read clustering (to remove random errors). The validity of this method is supported through formal statistical argument and analysis of control HCV and HIV data.While control data is an invaluable tool for assessing SNV calling, it is not always available. To overcome this limitation, I implemented the software GemSIM. GemSIM creates and uses empirically derived, sequence context based error models to emulate individual sequencing runs. Analysis of models from Illumina and Roche-454 data showed that error profiles vary even between different runs of the same technology, affecting SNV calling accuracy.Finally, I performed a pioneering application of deep sequencing to bacteria. Using matched samples, variants with frequencies down to 0.5% were detected in biofilms from Pseudomonas aeruginosa 18A (a clinical isolate) and PA01 (a lab strain). Biofilm microevolution was reproducible within a strain: PA01 biofilms featured rapid evolution and growth of PA01’s phage Pf4, with strong selection for c repressor mutants; while 18A biofilms featured selection of specific variants within the bacterial genome. Comparison with a Phaeobacter gallaciensis biofilm indicated that in general, adaptation within biofilms is typified by selection for a small number of non-synonymous variants within key genes involved in biofilm and competition related pathways.In conclusion, through careful analysis, experimental design, and the use of controls, sequence errors may be overcome, allowing deep sequencing to reveal the molecular evolution of microbes in unprecedented detail