Evidence is accumulating that noncoding RNAs and circRNA are involved in psoriasis; however, the competing endogenous RNA (ceRNA) mediated regulatory mechanisms in psoriasis are rarely reported. The research study aimed to comprehensively investigate the differences in the expression levels of circular RNA (circRNA), long non-coding RNA (lncRNA), microRNA (miRNA/miR), and mRNA in psoriasis. In addition, key lncRNA/circRNA-miRNA-mRNA-ceRNA interactions were screened using the GSE145305 microarray dataset from the Gene Expression Omnibus database. After data preprocessing, differentially expressed circRNAs (DECs), lncRNAs (DELs), miRNAs (DEMs), or genes (DEGs) were identified, and normal controls using the linear models for the microarray data method. A protein-protein interaction (PPI) network was constructed for DEGs based on protein databases, followed by a module analysis. The ceRNA network was constructed based on the interaction between miRNAs and mRNAs and lncRNAs/circRNAs and miRNAs. The present study identified that in the case of mRNA 10 genes are significantly down-regulated, 86 genes are significantly up-regulated and in the case of miRNA 48 are significantly down-regulated and 75 genes are significantly up-regulated between patients with psoriasis and controls. miRNA, mRNA, lncRNA, and circRNA target predictionswere made. Then combined construction of a ceRNA network using mRNA-miRNA-lncRNA and mRNAmiRNA-circRNA. The current research has employed the knowledge of bioinformatics tools and software to determine the hub module and PPI network. Taken together, these identified ceRNA interactions may be crucial targets for the treatment of psoriasis.

Sepsis is a complex, deadly, and difficult-to-diagnose disease characterized by anomalies in numerous life-threatening organ failures caused by an improper host response to an infecting organism such as bacteria, fungi, or viruses. Patient characteristics such as age and immunologic state, infection factors, and environmental factors such as nutritional status affect sepsis prognosis and make it difficult and a common cause of mortality. This project aimed to identifysepsis prognostic biomarkers by identifying significantly differentially expressed biomarkers across patient groups, then developing and evaluating a classification model that can help predict patients' prognosis. The project used input data consisting of 368 protein measurementsrepresented as Normalized Protein expression. These data have been preprocessed, split, and analyzed using the Wilcoxon rank-sum test to identify the significantly expressed biomarkers in each patient's subgroup, one in the ICU admission and six in the non-survived subgroups. These significantly expressed biomarkers were Volcano plotted, then integrated into different supervised and unsupervised multivariate statistical models. The best prognosis models for ICU admission were the KNN models based solely on either procalcitonin or C-reactive protein with AUCs of 1.00 (95% Cl: 1.00-1.00). The best prognosis model for the 28 days mortality was the KNN model of the tenascin-C with an AUC of 1.00 (95% Cl: 1.00-1.00). However, further studies are suggested using a larger sample size in order to lessen the likelihood of bias. Some of the identified significantly expressed biomarkers, procalcitonin, and CRP, could generate KNN models with high AUC that can be used to prognosis the ICU admission or the 28 days of mortality due to sepsis.

Shiga toxin-producing E. coli (STEC) are zoonotic pathogens that frequently colonise ruminants without them showing any symptoms. In humans, STEC cause diarrhoeal disease and occasionally leads to the life-threatening disease haemolytic-uraemic syndrome (HUS). In this study, the aim is to identify any genomic differences between Swedish STEC O157:H7 isolates that have caused HUS and isolates that did not, as well as between isolates taken from animals and isolates taken from humans. I constructed a pan-genome analysis pipeline and performed statistical analyses to find genes that differed between these groups. I also constructed a phylogenetic analysis pipeline to visualise any clustering of isolates based on different categories. The results from the phylogenetic analysis showed that the isolates tended to not form clear clusters based on their category. When comparing isolates from animals to isolates from humans, an elastic net regression analyses yielded a list of 23 genes that differed between them, while a statistical analysis using Scoary found 1854 genes. The genes found by the regression analysis consists largely of genes associated with metabolism, with other notable genes being transposases as well as two genes from the prp operon. Gene ontology analysis of the genes from Scoary showed that no particular molecular functions or biological processes stand out when compared to the background frequency of gene ontology terms. When comparing isolates that caused HUS against isolates that did not, no genes were found to be statistically significant. In order to find more conclusive results about the genomic differences between STEC in animals and humans, as well as between STEC that leads to HUS and STEC that does not, further studies are needed.

Campylobacter is the leading cause of gastroenteritis worldwide and in Sweden there areofficial programs for the surveillance of the bacteria. One important objective with foodbornepathogen surveillance is molecular typing. As typing based on whole genome sequencing datais becoming more common, knowledge on how to set up analysis pipelines is essential to avoidvariation in results. Here, typical whole genome sequencing pipelines are compared to areference genome at different analysis stages to optimize assembly quality and typing resultsusing cgMLST. The results show that read trimming is optimal to obtain high quality assemblieswith SPAdes as well as for improving cgMLST results compared to when no read trimming wasperformed before assembling with SPAdes. The opposite was shown for SKESA wheretrimming beforehand had negative effects on the results, most likely due to SKESA having builtin trimming properties. Additionally post assembly improvements had generally positive effects,however these effects were small.Tekni

Animal models have a long history of being used in research for the purpose of investigating biological processes and testing the effect of specific compounds on the functionality of biological processes. Different types of mice are used as animal models, most notably inbred and outbred strains. This study investigates the effect of certain priming conditions on the production of cytokines in wild mice and lab mice, using multivariate data analysis. This analytical study involves exploratory analysis, in the form of PCA, MANOVA and LDA, training of different classification models and their validation. Based on the conducted exploratory analysis, certain priming conditions (CD3CD28, CPG and PG) have been identified as clearly defined groups by PCA and LDA, in both wild mice and lab mice. MANOVA concluded that most of the variables tested are statistically significant in determining group association. Subsequent classification modeling determined that the Random Forest algorithm is the most accurate in predicting class, in both the wild and lab mice. The performed analysis has given insight into the major trends exhibited by the data, but further post-processing analysis could potentially extract more data. The results of this study could be used to further investigate the discovered pattern in the data or be supplemented by comparing additional mouse strains under the same experimental conditions.

The massive accumulation of omics data requires effective computational tools to analyze and interpret such data. Deep learning (DL), a branch of machine learning (ML) and artificial intelligence (AI), has shed light on these challengings and achieved great success in bioinformatics. However, the influence of preprocessing steps on DL model’s performance remains a critical aspect that requires thorough investigation. This study aims to investigates the effects of different combinations of preprocessing techniques and feature selection methods on the predictive performance of deep neural networks (DNN) on supervised tasks. For this purpose, four normalization methods, one transformation method, and two feature scaling methods were applied, in addition to two feature selection methods. This comprehensive analysis resulted in a total of 28 unique combinations, each representing a unique classifier. The experimental analysis was conducted using gene expression profiles from multiple cancer datasets. The result highlights the significance of preprocessing step in achieving optimal DNN performance, with notable variations observed across different datasets and preprocessing techniques. We identify a specific preprocessing workflow that improve DNN performance, and certain preprocessing choices that may lead to suboptimal model performance. In addition, we identify potential pitfalls and challenges associated with the data structure and class imbalance. This study contributes to the understanding of the effect of pre-processing steps and provides insights into which pre-processing steps work best and hence, improve the overall performance of DNN model and enables the development of more robust and accurate models.

Epigenomics is the study of modifications to the genetic material without changes to the DNA sequence, one such modification is methylation of nucleotides. DNA methylation is associated with gene regulation and is studied in a variety of fields such as cancer and ageing. Quality control is essential when designing research studies to ensure that the end result is not affected by poor quality data. In this study, the aim was to define robust quality parameters for whole methylome sequencing for Illumina next generation sequencing data. Three different library preparation protocols, all designed for methylation analysis, has been compared: Accel-NGS Methyl-Seq DNA library, NEB Next Enzymatic Methyl-seq and SPlinted Ligation Adapter Tagging. All samples were sequenced on the Illumina NovaSeq 6000 with paired-end 150 bp. An evaluation of alignment software was also included in the study. The nf-core methylseq pipeline version 1.6.1 was used to process all samples in the study. The pipeline was run multiple times with different settings depending on library type and software choice. Throughout the study, the parameters puc19, lambda and alignment rate showed consistency whereas overall methylation rate and coverage were affected by origin of sample material and study design. In conclusion, not all proposed quality parameters were suitable for general quality control since study design and origin of sample material have impact, but alignment rate and the controls puc19 and lambda shows great promise for general quality control. Future work to establish sample material specific thresholds for methylation rate is encouraged.

In this work we present a novel method to extract potential hub genes, transcription factors and regions with densely interconnected protein-protein-interaction networks from RNAseq data. To achieve this we deploy variational autoencoders, a generative machine learning framework, and extract the gene-wise reconstruction errors. This reconstruction error produced during training is considered as a measurement of impact for a gene on the transcriptome here.  The method can handle big datasets (3.5Gb and more) in reasonable time on computers for domestic usage without any gpu-acceleration. This circumstance allows users without access to large amounts of computational resources to also work with expression data of large size.  The final ranking based on reconstruction errors underlies less of a bias compared to most hub gene inference methods currently available. Also no prior gene regulatory network inference is required. However, the introduction of a bias can help to focus on certain genes of interest. Here we biased by using genes present in the STRING data base to also ease the following analysis.  Analysis of reconstruction error showed a tendency for genes with low reconstruction error to capture genes with central meaning to the data set used for training. In case of healthy cells this was genes associated with house keeping mechanisms and for breast cancer data those genes were associated to breast cancer. In breast cancer specific data we found for example a high frequency of HOX family members linked specifically to breast cancer. For data covering different types of cancer here the picture was broader and covered a wide range of genes associated with different types of cancer.  There also was a high enrichment of transcription factors present in the genes with low reconstruction error. Not only the regions with lowest reconstruction error will reveal a high enrichment for transcription factors, also other regions show transcription factor enrichment. Transcription factors from these other regions will differ regarding their correlation patterns.  Regions with low reconstruction error and/or a high transcription factor enrichment show a high PPI-enrichment and exhibit densely interconnected networks.

Sepsis is a dangerous and potentially fatal condition that has a mysterious origin, underscoring the significance of prompt and accurate diagnosis and treatment. Bacterial whole-genome sequencing, which is widely used in clinical microbiology, stands at the forefront of sequencing technologies, particularly to combat sepsis. The aim of this thesis is to improve sepsis treatment by examining the genetic characteristics and drug resistance patterns of the common sepsis-causing bacteria Pseudomonas and Proteus spp., by analyzing the whole-genome sequencing data of bacterial isolates using an in-house-developed pipeline. The result was compared with a commercial cloud-based platform from 1928 Diagnostic (Gothenburg, Sweden), as well as the results from a clinical laboratory. Using Illumina HiSeq X next-generation sequencing technology, whole-genome data from 88 isolates of Pseudomonas and Proteus spp. was obtained. The isolates were obtained during a prospective observational study of community-onset severe sepsis and septic shock in adults at Skaraborg Hospital in Sweden's western region. The collected isolates were characterized using approved laboratory techniques, such as phenotypic antibiotic susceptibility testing (AST) in accordance with EUCAST guidelines and species identification by MALDI-TOF MS analysis. The species identification result matched the phenotypic method, with the exception of two isolates from Pseudomonas samples and four isolates from Proteus samples. When benchmarking the in-house pipeline and 1928 platform for Pseudomonas spp., predicted 97% of the isolates were resistant to at least one class of the tested antibiotics, of which 94% shows multi-drug resistance. In phenotypes, 88% of the isolates had at least one antibiotic resistance future, of which 68% shows multi-drug resistance. The most prevalent sequence types (STs) identified were ST 3285 and ST111 (9.3%) and ST564 and ST17 (6.98%) each, and both pipelines accurately predicted the number of multilocus types. The in-house pipeline reported 9820 Pseudomonas virulence genes, with PhzB1, a metabolic factor, being the most common gene. It was discovered that there was a significant correlation between the virulence factor gene count and the multilocus sequence typing (MLST) (p = 0.00001). With a Simpson's Diversity Index of 0.98, the urine culture specimens showed the greatest ST diversity. Plasmids were detected in twelve samples (20.93%) in total. In general, this study provided a detailed description of the bacterial future for Pseudomonas and Proteus organisms using WGS data. This research shows the applicability of the in-house and 1928 pipelines in the identification of sepsis-causing organisms with accuracy. It also showed the need for an organized and easy-to-use international pipeline to implement and analyze WGS bacterial data and to compare it with laboratory results as needed.

Copy number variations (CNVs) cover approximately 4.9 - 9.5% of the human genome. CNVs are involved in both the development of disease and evolutionary adaptions. CNVs play an important part in the development and progression of multiple cardiovascular diseases. CNV calling is traditionally performed with cromosomal microarray (CMA). The advantage of next generation sequensing (NGS) instead of CMA include higher resolution, lower cost and higher sensitivity in detecting smaller CNVs. CNV calling with NGS is connected to a high number of false positives. In this study three different CNV-calling tools for clinical exome sequencing data were evaluated; CoNIFER, CONTRA and ExomeDepth. To further decrease the false positive rate and decrease the hands-on analysis time a variant filter for ExomeDepth was developed and evaluated. However, CNV-calling with clinical exome data is still challenging due to low coverage.