Biology subjects

Halla-aho, V.

Publications and source records attributed to Halla-aho, V..

4 recordsLinked to original sources

bioRxiv2021

Probabilistic modeling methods for cell-free DNA methylation based cancer classification

BackgroundcfMeDIP-seq is a low-cost method for determining the DNA methylation status of cell-free DNA and it has been successfully combined with statistical methods for accurate cancer diagnostics. We investigate the diagnostic classification aspect by applying statistical tests and dimension reduction techniques for feature selection and probabilistic modeling for the cancer type classification, and we also study the effect of sequencing depth. MethodsWe experiment with a variety of statistical methods that use different feature selection and feature extraction methods as well as probabilistic classifiers for diagnostic decision making. We test the (moderated) t-tests and the Fishers exact test for feature selection, principal component analysis (PCA) as well as iterative supervised PCA (ISPCA) for feature generation, and GLMnet and logistic regression methods with sparsity promoting priors for classification. Probabilistic programming language Stan is used to implement Bayesian inference for the probabilistic models. Results and conclusionsWe compare overlaps of differentially methylated genomic regions as chosen by different feature selection methods, and evaluate probabilistic classifiers by evaluating the area under the receiver operating characteristic (AUROC) scores on discovery and validation cohorts. While we observe that many methods perform equally well as, and occasionally considerably better than, GLMnet that was originally proposed for cfMeDIP-seq based cancer classification, we also observed that performance of different methods vary across sequencing depths, cancer types and study cohorts. Overall, methods that seem robust and promising include Fishers exact test and ISPCA for feature selection as well as a simple logistic regression model with the number of hyper and hypomethylated regions as features.

bioinformatics↗

bioRxiv2021

Permutation-based significance analysis reduces the type 1 error rate in bisulfite sequencing data analysis of human umbilical cord blood samples

BackgroundDNA methylation patterns are largely established in-utero and might mediate the impacts of in-utero conditions on later health outcomes. Associations between perinatal DNA methylation marks and pregnancy-related variables, such as maternal age and gestational weight gain, have been earlier studied with methylation microarrays, which typically cover less than 2 % of human CpG sites. To detect such associations outside these regions, we chose the bisulfite sequencing approach. MethodsWe collected and curated all available clinical data on 200 newborn infants; whose umbilical cord blood samples were analyzed with the reduced representation bisulfite sequencing (RRBS) method. A generalized linear mixed effects model was fit for each high coverage CpG site, followed by spatial and multiple testing adjustment of P values to identify differentially methylated cytosines (DMCs) and regions (DMRs) associated with clinical variables such as maternal age, mode of delivery, and birth weight. Type 1 error rate was then evaluated with a permutation analysis. ResultsWe discovered a strong inflation of spatially adjusted P values through the permutation analysis, which we then applied for empirical type 1 error control. Based on empirically estimated significance thresholds, very little differential methylation was associated with any of the studied clinical variables, other than sex. With this analysis workflow, the sex-associated differentially methylated regions were highly reproducible across studies, technologies, and statistical models. ConclusionsThe inflation of P values was caused by a common method for spatial adjustment and DMR detection, implemented in tools comb-p and RADMeth. With standard significance thresholds, type 1 error rates were high with both these implementations, across alternative parameter settings and analysis strategies. We conclude that comb-p and RADMeth are convenient methods for the detection of differentially methylated regions, but the statistical significance should either be determined empirically or before the spatial adjustment. Our RRBS data analysis workflow is available in https://github.com/EssiLaajala/RRBS_workflow.

bioinformatics↗

bioRxiv2020

LuxHS: DNA methylation analysis with spatially varying correlation structure

Bisulfite sequencing (BS-seq) is a popular method for measuring DNA methylation in basepair-resolution. Many BS-seq data analysis tools utilize the assumption of spatial correlation among the neighboring cytosines methylation states. While being a fair assumption, most existing methods leave out the possibility of deviation from the spatial correlation pattern. Our approach builds on a method which combines a generalized linear mixed model (GLMM) with a likelihood that is specific for BS-seq data and that incorporates a spatial correlation for methylation levels. We propose a novel technique using a sparsity promoting prior to enable cytosines deviating from the spatial correlation pattern. The method is tested with both simulated and real BS-seq data and compared to other differential methylation analysis tools.

bioinformatics↗

bioRxiv2019

LuxUS: Detecting differential DNA methylation using generalized linear mixed model with spatial correlation structure

AO_SCPLOWBSTRACTC_SCPLOWO_ST_ABSMotivationC_ST_ABSDNA methylation is an important epigenetic modification, which has multiple functions. DNA methylation and its connections to diseases have been extensively studied in recent years. It is known that DNA methylation levels of neighboring cytosines are correlated and that differential DNA methylation typically occurs rather as regions instead of individual cytosine level. ResultsWe have developed a generalized linear mixed model, LuxUS, that makes use of the correlation between neighboring cytosines to facilitate analysis of differential methylation. LuxUS implements a likelihood model for bisulfite sequencing data that accounts for experimental variation in underlying biochemistry. LuxUS can model both binary and continuous covariates, and mixed model formulation enables including replicate and cytosine random effects. Spatial correlation is included to the model through a cytosine random effect correlation structure. We show with simulation experiments that by utilizing the spatial correlation we gain more power to the statistical testing of differential DNA methylation. Results with real bisulfite sequencing data set show that LuxUS is able to detect biologically significant differentially methylated cytosines. AvailabilityThe tool is available at https://github.com/hallav/LuxUS. Supplementary informationSupplementary data are available at bioRxiv.

bioinformatics↗