Comprehensive characterisation of the mutational landscapes of solid tumours is a multistep process involving the collection of suitable samples, the extraction of nucleic acids and the preparation of these materials for mutational analyses. In this thesis, I aimed to develop a streamlined process for the analysis of colorectal cancer (CRC) patient samples in order to identify novel mutations that hallmark the development of advanced disease. Papers I and II outline a technique for serial extraction of nucleic acids from frozen tissue that we developed and subsequently implemented on a robotic platform to enable high-throughput processing. The extracted nucleic acids were validated in downstream processes relevant for genetic analyses, including traditional Sanger and next generation sequencing  techniques. In Paper III, we developed a genotyping method based on multiplex ligation-dependent genome amplification. The method was designed such that InDel polymorphisms of between 30 and 70 % prevalence in a European population were selected and amplified in a multiplex PCR assay. DNA from 24 patient-matched colorectal tumour and normal tissues was genotyped and paired with a high match probability. In Paper IV, we performed targeted resequencing of 107 primary CRCs, of which approximately half developed metastatic disease or had distant metastases at the time of diagnosis. We chose to analyse 676 genes based on their involvement in key signalling pathways in CRC. We found an enrichment of mutations in the Eph receptor tyrosine kinase gene family in metastatic patients, indicating a potential role for these genes in CRC metastasis. This thesis outlines a series of procedures that can be employed in a high-throughput setting for the analysis of solid tumours. We applied these methods to the analysis of colorectal tumours and propose a link between novel somatic mutations and metastatic disease.

Schizophrenia demonstrates high heritability in part accounted for by common simple nucleotide variants (SNV), rare copy number variants (CNV) and, most recently, rare SNVs Although heritability explained by rare SNVs and CNVs is small compared to that explained by common SNVs, rare SNVs in functional sequences may identify specific disease mechanisms. However, current exome methods do not capture a large proportion of potentially functional bases where rare variation may impact disease risk: as much as two-thirds of conserved sequences lie outside the exome in non-coding regions of cross-species evolutionary constraint. We reasoned that the candidate loci from the Psychiatric Genomics Consortium Phase 1 (PGC-1) schizophrenia study represent good target loci to test for the impact of rare SNVs in non-coding constrained regions. We developed custom reagents to capture mammalian constrained non-coding regions, exons, and 5’- and 3’-untranslated regions (UTRs) in the 12 PGC-1 loci for pooled sequencing in 912 cases and 936 controls. Compared to our coding targets, our noncoding targets contain substantially more highly conserved bases (46,412 vs. 31,609) and variants (390 vs. 193). Using C-alpha to detect excess variance due to aggregate risk increasing or decreasing rare SNV effects, we identified signals attributable to alleles with MAF < 0.1% in both coding sequences and in functional non-coding sequences, including variants within ENCODE transcription factor binding sites, DNase hypersensitive regions, and histone modification sites in neuronal cell lines. We also observed significant excess risk-altering variation in the CUB domain of CSMD1, a gene expressed in the developing central nervous system. These results support the hypothesis that common and rare variants in the same loci contribute to schizophrenia risk, but highlight the need to expand capture strategies in order to detect trait-relevant sequence variation in a broader set of functional sequences.

Moderní výzkum v environmentální mikrobiologii využívá k popisu mikrobiálních komunit genomická data, především sekvenaci DNA. Oblast, která zkoumá veškerý genetický materiál přítomný v environmentálním vzorku, se nazývá metagenomika. Tato doktorská práce se zabývá metagenomikou z pohledu bioinformatiky, která je nenahraditelná při výpočetním zpracování dat. V teoretické části práce jsou popsány dva základní přístupy metagenomiky, včetně jejich základních principů a slabin. První přístup, založený na cíleném sekvenování, je dobře rozpracovanou oblastí s velkou řadou bioinformatických technik. Přesto mohou být metody pro porovnávání vzorků z několika prostředí podstatně vylepšeny. Přístup představený v této práci používá unikátní transformaci dat do podoby bipartitního grafu, kde je jedna partita tvořena taxony a druhá vzorky, případně různými prostředími. Takový graf plně reflektuje kvalitativní i kvantitativní složení analyzované mikrobiální sítě. Umožňuje masivní redukci dat pro jednoduché vizualizace bez negativních vlivů na automatickou detekci komunit, která dokáže odhalit shluky podobných vzorků a jejich typických mikrobů. Druhý přístup využívá sekvenace celého metagenomu. Tato strategie je novější a příslušející bioinformatické nástroje jsou méně propracované. Hlavní výzvou přitom zůstává rychlá klasifikace sekvencí, v metagenomice označovaná jako „binning“. Metoda představená v této práci využívá přístupu zpracování genomických signálů. Tato unikátní metodologie byla navržena na základě podrobné analýzy redundance genetické informace uložené v genomických signálech. Využívá transformace znakových sekvencí do několika variant fázových signálů. Navíc umožňuje přímé zpracování dat ze sekvenace nanopórem v podobě nativních proudových signálů.

In breast cancer, the genetic profiling of circulating cell-free DNA (cfDNA) from blood plasma was shown to have good potential for clinical use. In contrast, only a few studies were performed investigating urinary cfDNA. In this pilot study, we analyzed plasma-derived and matching urinary cfDNA samples obtained from 15 presurgical triple-negative breast cancer patients. We used a targeted next-generation sequencing approach to identify and compare genetic alterations in both body fluids. The cfDNA concentration was higher in urine compared to plasma, but there was no significant correlation between matched samples. Bioinformatical analysis revealed a total of 3339 somatic breast-cancer-related variants (VAF ≥ 3%), whereof 1222 vs. 2117 variants were found in plasma-derived vs. urinary cfDNA, respectively. Further, 431 shared variants were found in both body fluids. Throughout the cohort, the recovery rate of plasma-derived mutations in matching urinary cfDNA was 47% and even 63% for pathogenic variants only. The most frequently occurring pathogenic and likely pathogenic mutated genes were NF1, CHEK2, KMT2C and PTEN in both body fluids. Notably, a pathogenic CHEK2 (T519M) variant was found in all 30 samples. Taken together, our results indicated that body fluids appear to be valuable sources bearing complementary information regarding the genetic tumor profile.

This thesis describes the development and use of a new class of molecular tools called Selector probes, and its potential for investigations of genetic variation. The Selector technology provides multiplex amplification of targeted DNA sequences with a high specificity, and an enrichment factor in the same order of magnitude as PCR. A common feature in this thesis work is to focus the analysis on DNA regions of interest. For example, this technique can be implemented in analysing candidate regions found by whole genome studies that need validation (global to local analysis), and applications requiring detection of rare alleles (common to rare allele), important in for example cancer samples. An assay is presented that allows for fast and simple quantification of relative copy-number variations. The method was proven to be able to detect aneuploidy in chromosome 13, 18, 21 and X, with a resolution enough to distinguish between 4 and 5 copies. The method was successfully applied to solve a biological question regarding a copy-number variation, that explains the Ridge phenotype typical for the dog bread Rhodesian Ridgebacks. The Selector strategy was able to detect and map a tandem duplication with a size of 133 kb, which was characterized with base-pair resolution. A readout platform that facilitates simultaneous digital quantitative analysis of a large numbers of biomolecules is further introduced. The work involves arraying amplified product from successful selection and decoding each molecule by hybridization of fluorophore labeled oligonucleotides. Finally, a genome partitioning method which is applied upstream of next generation sequencing platforms is presented. It is shown that the method provides successful enrichment with 98 % coverage and 94 % specificity and high enrichment uniformity. The technique was applied for mutation analysis of 26 cancer-related genes in tumor cell-lines and tissue.