| 41. |
- Borgmästars, Emmy, 1990-
(författare)
-
In search of early biomarkers in pancreatic ductal adenocarcinoma using multi-omics and bioinformatics
- 2022
-
Doktorsavhandling (övrigt vetenskapligt/konstnärligt)abstract
- Background: Pancreatic ductal adenocarcinoma (PDAC) is a very aggressive malignancy with a 5-year survival of 10 %. Surgery is the only curative treatment. Unfortunately, few patients are eligible for surgery due to late detection. Thus, we need ways to detect the disease at an earlier stage and for that good screening biomarkers could be used. Previous studies have analyzed circulating analytes in prospective studies to identify early PDAC signals. One such class is microRNAs (miRNAs). MicroRNAs are non-coding RNAs of around 22 nucleotides that act as post- transcriptional regulators by interaction with messenger RNAs (mRNAs). The function of a miRNA can be elucidated by target prediction, to identify its potential targets, followed by enrichment analysis of the predicted targets. Challenges with this approach includes a lot of false positives being generated and that miRNAs can perform their role in a tissue- or disease-specific manner. Other classes of analytes that have previously been studied in prospective PDAC cohorts are metabolites and proteins. Aims: This thesis has three aims. First, to build a miRNA functional analysis pipeline with correlation support between miRNA and its predicted target genes. Second, to identify potential circulating biomarkers for early detection of PDAC using multi-omics. Third, to identify potential prognostic metabolites in a prospective PDAC cohort.Methods: We used publicly available data from the cancer genome atlas-pancreatic adenocarcinoma (TCGA-PAAD) and pre-diagnostic plasma samples from the Northern Sweden Health and Disease Study. We built a pipeline in R including miRNA, mRNA, and protein expression data from TCGA-PAAD for in silico miRNA functional analysis. Pre- diagnostic plasma samples from future PDAC patients as well as matched healthy controls were analyzed using multi- omics. Tissue polypeptide specific antigen (TPS) was analyzed by enzyme linked immunosorbent assay in 267 future PDAC samples and 320 healthy controls. Metabolomics and clinical biomarkers (carbohydrate antigen (CA) 19-9, carcinoembryonic antigen (CEA), and CA 15-3) were profiled in 100 future PDAC samples and 100 healthy controls using liquid chromatography-mass spectrometry (MS), gas chromatography-MS, and multi-plex technology. Of these, a subset of 39 future PDAC patients and 39 healthy controls were profiled for 2083 microRNAs using targeted sequencing and 644 proteins using proximity extension assays. Circulating levels of multi-omics analytes were analyzed using conditional or unconditional logistic regression. Least absolute shrinkage and selection operator (LASSO) in combination with 500 bootstrap iterations identified the most informative variables. The prognostic value of metabolites was assessed using cox regression. Multi-omics factor analysis (MOFA) and data integration analysis for biomarker discovery using latent components (DIABLO) were used for multi-omics integration analyses.Results: An automated pipeline was built consisting of 1) miRNA target prediction, 2) correlation analyses between miRNA and its targets on mRNA and protein expression levels, and 3) functional enrichment of correlated targets to identify enriched Kyoto encyclopedia of genes and genomes (KEGG) pathways and gene ontology (GO) terms for a specific miRNA. The pipeline was run for all microRNAs (~700) detected in the TCGA-PAAD cohort. These results can be downloaded from a shiny app (https://emmbor.shinyapps.io/mirfa/). TPS was not altered in pre-diagnostic PDAC patients up to 24 years prior to diagnosis, but increased at diagnosis (OR = 1.03, 95 % CI: 1.01-1.05). Internal area under curves of 0.74, 0.80, and 0.88 were achieved for five metabolites, two proteins, and two miRNAs that were selected by LASSO and bootstrap iterations, in combination with CA 19-9. Neither MOFA nor DIABLO separated well between future PDAC cases and healthy controls. Conclusions: Our bioinformatics pipeline for in silico functional analysis of microRNAs successfully identifies enriched KEGG pathways and GO terms for miRNA isoforms. The investigated plasma samples are heterogeneous, but among the analyzed variables, we identified five metabolites, two proteins, and two microRNAs with highest potential for early PDAC detection. CA 19-9 levels increased closer to diagnosis. We identified five fatty acids that could be studied in a diagnostic PDAC cohort as prognostic biomarkers.
|
|
| 42. |
- Bornelöv, Susanne, et al.
(författare)
-
Visualization of Rules in Rule-Based Classifiers
- 2012
-
Ingår i: INTELLIGENT DECISION TECHNOLOGIES (IDT'2012), VOL 1. - Berlin, Heidelberg : Springer Berlin Heidelberg. - 9783642299773 ; , s. 329-338
-
Konferensbidrag (refereegranskat)abstract
- Interpretation and visualization of the classification models are important parts of machine learning. Rule-based classifiers often contain too many rules to be easily interpreted by humans, and methods for post-classification analysis of the rules are needed. Here we present a strategy for circular visualization of sets of classification rules. The Circos software was used to generate graphs showing all pairs of conditions that were present in the rules as edges inside a circle. We showed using simulated data that all two-way interactions in the data were found by the classifier and displayed in the graph, although the single attributes were constructed to have no correlation to the decision class. For all examples we used rules trained using the rough set theory, but the visualization would by applicable to any sort of classification rules. This method for rule visualization may be useful for applications where interaction terms are expected, and the size of the model limits the interpretability.
|
|
| 43. |
- Broberg, Karin, et al.
(författare)
-
Evaluation of 92 cardiovascular proteins in dried blood spots collected under field-conditions : Off-the-shelf affinity-based multiplexed assays work well, allowing for simplified sample collection
- 2021
-
Ingår i: BioEssays. - : Wiley. - 0265-9247 .- 1521-1878. ; 43:9
-
Tidskriftsartikel (refereegranskat)abstract
- Workplace-collected blood spots deposited on filter paper were analysed with multiplexed affinity-based protein assays and found to be suitable for proteomics analysis. The protein extension assay (PEA) was used to characterize 92 proteins using 1.2 mm punches in repeated samples collected from 20 workers. Overall, 97.8% of the samples and 91.3% of the analysed proteins passed quality control. Both within and between spot correlations using six replicates from the same individual were above 0.99, suggesting that comparable levels are obtained from multiple punches from the same spot and from consecutive spots. Protein levels from dried blood and wet serum from the same individuals were compared and the majority of the analysed proteins were found to be significantly correlated. These results open up for simplified sample collection of blood in field conditions for proteomic analysis, but also highlight that not all proteins can be robustly measured from dried whole blood.
|
|
| 44. |
- Chen, Dan, et al.
(författare)
-
Genome-wide Association Study of Susceptibility Loci for Cervical Cancer
- 2013
-
Ingår i: Journal of the National Cancer Institute. - : Oxford University Press (OUP). - 0027-8874 .- 1460-2105. ; 105:9, s. 624-633
-
Tidskriftsartikel (refereegranskat)abstract
- Background Cervical carcinoma has a heritable genetic component, but the genetic basis of cervical cancer is still not well understood. Methods We performed a genome-wide association study of 731 422 single nucleotide polymorphisms (SNPs) in 1075 cervical cancer case subjects and 4014 control subjects and replicated it in 1140 case subjects and 1058 control subjects. The association between top SNPs and cervical cancer was estimated by odds ratios (ORs) and 95% confidence intervals (CIs) with unconditional logistic regression. All statistical tests were two-sided. Results Three independent loci in the major histocompatibility complex (MHC) region at 6p21.3 were associated with cervical cancer: the first is adjacent to the MHC class I polypeptide-related sequence A gene (MICA) (rs2516448; OR = 1.42, 95% CI = 1.31 to 1.54; P = 1.6 x 10(-18)); the second is between HLA-DRB1 and HLA-DQA1 (rs9272143; OR = 0.67, 95% CI = 0.62 to 0.72; P = 9.3 x 10(-24)); and the third is at HLA-DPB2 (rs3117027; OR=1.25, 95% CI = 1.15 to 1.35; P = 4.9 x 10(-8)). We also confirmed previously reported associations of B*0702 and DRB1*1501-DQB1*0602 with susceptibility to and DRB1*1301-DQA1*0103-DQB1*0603 with protection against cervical cancer. The three new loci are statistically independent of these specific human leukocyte antigen alleles/haplotypes. MICA encodes a membrane-bound protein that acts as a ligand for NKG2D to activate antitumor effects. The risk allele of rs2516448 is in perfect linkage disequilibrium with a frameshift mutation (A5.1) of MICA, which results in a truncated protein. Functional analysis shows that women carrying this mutation have lower levels of membrane-bound MICA. Conclusions Three novel loci in the MHC may affect susceptibility to cervical cancer in situ, including the MICA-A5.1 allele that may cause impaired immune activation and increased risk of tumor development.
|
|
| 45. |
- Chen, Dan, et al.
(författare)
-
Pathway analysis of cervical cancer genome-wide association study highlights the MHC region and pathways involved in response to infection
- 2014
-
Ingår i: Human Molecular Genetics. - : Oxford University Press (OUP). - 0964-6906 .- 1460-2083. ; 23:22, s. 6047-6060
-
Tidskriftsartikel (refereegranskat)abstract
- Cervical cancer is caused by infection with human papillomavirus (HPV). A genome-wide association study (GWAS) has identified several susceptibility loci for cervical cancer, but they explain only a small fraction of cervical cancer heritability. Other variants with weaker effect may be missed due to the stringent significance threshold. To identify important pathways in cervical carcinogenesis, we performed a two-stage pathway analysis in two independent GWASs in the Swedish population, using the single-nucleotide polymorphism (SNP) ratio test. The 565 predefined pathways from Kyoto Encyclopedia of Genes and Genomes and BioCarta databases were systematically evaluated in the discovery stage (1034 cases and 3948 controls with 632 668 SNPs) and the suggestive pathways were further validated in the replication stage (616 cases and 506 controls with 341 358 SNPs). We found 12 pathways that were significant in both stages, and these were further validated using set-based analysis. For 10 of these pathways, the effect was mainly due to genetic variation within the major histocompatibility complex (MHC) region. In addition, we identified a set of novel candidate genes outside the MHC region in the pathways denoted ‘Staphylococcus aureus infection’ and ‘herpes simplex infection’ that influenced susceptibility to cervical cancer (empirical P = 4.99 × 10−5 and 4.99 × 10−5 in the discovery study; empirical P = 8.98 × 10−5 and 0.009 in the replication study, respectively). Staphylococcus aureus infection may evoke an inflammatory response that inadvertently enhances malignant progression caused by HPV infection, and Herpes simplex virus-2 infection may act in conjunction with HPV infection to increase the risk of cervical carcinoma development. These findings provide new insights into the etiology of cervical cancer.
|
|
| 46. |
- Chen, Dan, et al.
(författare)
-
Pooled analysis of genome-wide association studies of cervical intraepithelial neoplasia 3 (CIN3) identifies a new susceptibility locus
- 2016
-
Ingår i: Oncotarget. - : Impact Journals, LLC. - 1949-2553. ; 7:27, s. 42216-42224
-
Tidskriftsartikel (refereegranskat)abstract
- Recent genome-wide association studies (GWASs) in subjects of European descent have identified associations between cervical cancer risk and three independent loci as well as multiple classical human leukocyte antigen (HLA) alleles at 6p21.3. To search for novel loci associated with development of cervical cancer, we performed a pooled analysis of data from two GWASs by imputing over 10 million genetic variants and 424 classical HLA alleles, for 1,553 intraepithelial neoplasia 3 (CIN3), 81 cervical cancer and 4,442 controls from the Swedish population. Notable findings were validated in an independent study of 961 patients (827 with CIN3 and 123 with cervical cancer) and 1,725 controls. Our data provided increased support for previously identified loci at 6p21.3 (rs9271898, P = 1.2 x 10(-24); rs2516448, 1.1 x 10(-15); and rs3130196, 2.3 x 10(-9), respectively) and also confirmed associations with reported classical HLA alleles including HLA-B*07:02, -B*15:01, -DRB1*13:01, -DRB1*15:01, -DQA1*01:03, -DQB1*06:03 and -DQB1*06:02. In addition, we identified and subsequently replicated an independent signal at rs73730372 at 6p21.3 (odds ratio = 0.60, 95% confidence interval = 0.54-0.67, P = 3.0 x 10(-19)), which was found to be an expression quantitative trait locus (eQTL) of both HLA-DQA1 and HLA-DQB1. This is one of the strongest common genetic protective variants identified so far for CIN3. We also found HLA-C*07:02 to be associated with risk of CIN3. The present study provides new insights into pathogenesis of CIN3.
|
|
| 47. |
- Cui, Tao, 1982-, et al.
(författare)
-
Invasive cervical tumors with high and low HPV titer represent molecular subgroups with different disease etiology
- 2019
-
Ingår i: Carcinogenesis. - : Oxford University Press (OUP). - 0143-3334 .- 1460-2180. ; 40:2, s. 269-278
-
Tidskriftsartikel (refereegranskat)abstract
- Invasive cervical cancer (ICC) with very low titer of high-risk human papillomavirus (HPV) has worse clinical outcome than cases with high titer, indicating a difference in molecular etiology. Fresh-frozen ICC tumors (n = 49) were classified into high- and low-HPV-titer cases using real-time PCR-based HPV genotyping. The mutation spectra were studied using the AmpliSeq Comprehensive Cancer Panel and the expression profiles using total RNA sequencing, and the results were validated using the AmpliSeq Transcriptome assay. HPV DNA genotyping and RNA sequencing showed that 16.6% of ICC tumors contained very low levels of HPV DNA and HPV transcripts. Tumors with low HPV levels had more mutations with a high allele frequency and fewer mutations with low allele frequency relative to tumors with high HPV titer. A number of genes showed significant expression differences between HPV titer groups, including genes with somatic mutations. Gene ontology and pathway analyses implicated the enrichment of genes involved in DNA replication, cell cycle control and extracellular matrix in tumors with low HPV titer. The results indicate that in low titer tumors, HPVs act as trigger of cancer development whereas somatic mutations are clonally selected and become drivers of the tumor development process. In contrast, in tumors with high HPV titer the expression of HPV oncoproteins plays a major role in tumor development and the many low frequency somatic mutations represent passengers. This putative subdivision of invasive cervical tumors may explain the higher radiosensitivity of ICC tumors with high HPV titer and thereby have consequences for clinical management.
|
|
| 48. |
- Draminski, Michal, et al.
(författare)
-
Monte Carlo feature selection for supervised classification
- 2008
-
Ingår i: Bioinformatics. - : Oxford University Press (OUP). - 1367-4803 .- 1367-4811. ; 24:1, s. 110-117
-
Tidskriftsartikel (refereegranskat)abstract
- MOTIVATION: Pre-selection of informative features for supervised classification is a crucial, albeit delicate, task. It is desirable that feature selection provides the features that contribute most to the classification task per se and which should therefore be used by any classifier later used to produce classification rules. In this article, a conceptually simple but computer-intensive approach to this task is proposed. The reliability of the approach rests on multiple construction of a tree classifier for many training sets randomly chosen from the original sample set, where samples in each training set consist of only a fraction of all of the observed features. RESULTS: The resulting ranking of features may then be used to advantage for classification via a classifier of any type. The approach was validated using Golub et al. leukemia data and the Alizadeh et al. lymphoma data. Not surprisingly, we obtained a significantly different list of genes. Biological interpretation of the genes selected by our method showed that several of them are involved in precursors to different types of leukemia and lymphoma rather than being genes that are common to several forms of cancers, which is the case for the other methods.
|
|
| 49. |
|
|
| 50. |
- Ek, Weronica E, et al.
(författare)
-
Genetic variants influencing phenotypic variance heterogeneity
- 2018
-
Ingår i: Human Molecular Genetics. - : OXFORD UNIV PRESS. - 0964-6906 .- 1460-2083. ; 27:5, s. 799-810
-
Tidskriftsartikel (refereegranskat)abstract
- Most genetic studies identify genetic variants associated with disease risk or with the mean value of a quantitative trait. More rarely, genetic variants associated with variance heterogeneity are considered. In this study, we have identified such variance single-nucleotide polymorphisms (vSNPs) and examined if these represent biological gene x gene or gene x environment interactions or statistical artifacts caused by multiple linked genetic variants influencing the same phenotype. We have performed a genome-wide study, to identify vSNPs associated with variance heterogeneity in DNA methylation levels. Genotype data from over 10 million single-nucleotide polymorphisms (SNPs), and DNA methylation levels at over 430 000 CpG sites, were analyzed in 729 individuals. We identified vSNPs for 7195 CpG sites (P < 9.4 x 10(-11)). This is a relatively low number compared to 52 335 CpG sites for which SNPs were associated with mean DNA methylation levels. We further showed that variance heterogeneity between genotypes mainly represents additional, often rare, SNPs in linkage disequilibrium (LD) with the respective vSNP and for some vSNPs, multiple low frequency variants co-segregating with one of the vSNP alleles. Therefore, our results suggest that variance heterogeneity of DNA methylation mainly represents phenotypic effects by multiple SNPs, rather than biological interactions. Such effects may also be important for interpreting variance heterogeneity of more complex clinical phenotypes.
|
|
