Carel F. W. Peeters
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Carel F. W. Peeters has a total of 124 co-authors. Here are the top co-authors by number of shared publications:
- Wiesje Maria van der Flier
Neurodegeneration, Amsterdam Neuroscience, Amsterdam, Netherlands
5 shared publications - Charlotte Elisabeth Teunissen
Neurochemistry Laboratory, Department of Laboratory Medicine, Amsterdam UMC, Amsterdam, The Netherlands
4 shared publications - Wessel N. van Wieringen
Methodology, Amsterdam Public Health, Amsterdam, The Netherlands
4 shared publications Georgios Bartzis
Mathematical and Statistical Methods group (Biometris), Wageningen University and Research, Wageningen, The Netherlands
3 shared publications- John Cornelis Van Swieten
Department of Neurology, Erasmus University Medical Center, Rotterdam, the Netherlands
3 shared publications - Lieke H. Meeter
Department of Neurology, Alzheimer Center, Erasmus MC University Medical Center, Dr. Molewaterplein 40, 3000 CA, Rotterdam, The Netherlands
3 shared publications
Latest Publications
- 8 June 2026
- Published by The R Foundation
HAMMER: High-Dimensional Factor-Analytic Representation Modeling and Metrics
The goal of 'HAMMER' is to provide factor analytic representation learningand associated determinacy metrics for very-high-dimensional data. It projects high-dimensional data onto low-dimensional generative latent sources and assesses the uncertainty in the projection. The projection is distribution-free, scale-equivariant, and efficient.
- 8 June 2026
- Published by The R Foundation
HAMMER: High-Dimensional Factor-Analytic Representation Modeling and Metrics
The goal of 'HAMMER' is to provide factor analytic representation learningand associated determinacy metrics for very-high-dimensional data. It projects high-dimensional data onto low-dimensional generative latent sources and assesses the uncertainty in the projection. The projection is distribution-free, scale-equivariant, and efficient.
- Open Access
- 27 August 2025
- Published by Springer Nature in Molecular Neurodegeneration
- Vol. 20 (1) , 1-17
Large-scale CSF proteome profiling identifies biomarkers for accurate diagnosis of frontotemporal dementia
Background: Diagnosis of Frontotemporal dementia (FTD) and its specific underlying neuropathologies (frontotemporal lobar degeneration; FTLD-Tau and FTLD-TDP) are challenging, and thus, fluid biomarkers are needed to improve diagnostic accuracy. Methods: We used proximity extension assays to analyze 665 proteins in cerebrospinal fluid (CSF) samples from a multicenter cohort, which included patients with FTD ( n = 189), Alzheimer’s Disease dementia (AD; n = 232), and cognitively unimpaired individuals ( n = 196). In a subset, FTLD neuropathology was determined based on phenotype or genotype (FTLD-Tau = 87 and FTLD-TDP = 67). Differences in protein expression profiles were analyzed using nested linear models. Penalized generalized linear modeling was used to identify classification protein panels, which were translated to custom multiplex assays and validated in two clinical cohorts (cohort 1: n = 161; cohort 2: n = 162), one autopsy-confirmed cohort ( n = 100), and one genetic cohort ( n = 55). Results: Forty-three proteins were differentially regulated in FTD compared to controls and AD, reflecting axon development, regulation of synapse assembly, and cell-cell adhesion mediator activity pathways. Classification analysis identified a 14- and 13-CSF protein panel that discriminated FTD from controls (FTD diagnostic panel, AUC: 0.96) or AD (FTD differential diagnostic panel, AUC: 0.91). Custom multiplex panels confirmed the strong discriminative performancen between FTD and controls (AUCs > 0.96) and between FTD and AD (AUCs > 0.88) across three validation cohorts, including one with autopsy confirmation (AUCs > 0.90). Validation in genetic FTD (including C9orf72 , GRN , and MAPT mutation carriers) revealed high accuracy of the FTD diagnostic panel in identifying both the presymptomatic (AUCs > 0.95) and symptomatic (AUC: 1) stages. Six proteins were differentially regulated between FTLD-TDP and FTLD-Tau. However, a reproducible classification model could not be generated (AUC: 0.80). Conclusions: Overall, this study introduces novel FTD-specific biomarker panels with potential use in diagnostic settings.
- 27 August 2025
- Published by Springer Nature in Molecular Neurodegeneration
- Vol. 20 (1) , 1-17
Open AccessLarge-scale CSF proteome profiling identifies biomarkers for accurate diagnosis of frontotemporal dementia
Background: Diagnosis of Frontotemporal dementia (FTD) and its specific underlying neuropathologies (frontotemporal lobar degeneration; FTLD-Tau and FTLD-TDP) are challenging, and thus, fluid biomarkers are needed to improve diagnostic accuracy. Methods: We used proximity extension assays to analyze 665 proteins in cerebrospinal fluid (CSF) samples from a multicenter cohort, which included patients with FTD ( n = 189), Alzheimer’s Disease dementia (AD; n = 232), and cognitively unimpaired individuals ( n = 196). In a subset, FTLD neuropathology was determined based on phenotype or genotype (FTLD-Tau = 87 and FTLD-TDP = 67). Differences in protein expression profiles were analyzed using nested linear models. Penalized generalized linear modeling was used to identify classification protein panels, which were translated to custom multiplex assays and validated in two clinical cohorts (cohort 1: n = 161; cohort 2: n = 162), one autopsy-confirmed cohort ( n = 100), and one genetic cohort ( n = 55). Results: Forty-three proteins were differentially regulated in FTD compared to controls and AD, reflecting axon development, regulation of synapse assembly, and cell-cell adhesion mediator activity pathways. Classification analysis identified a 14- and 13-CSF protein panel that discriminated FTD from controls (FTD diagnostic panel, AUC: 0.96) or AD (FTD differential diagnostic panel, AUC: 0.91). Custom multiplex panels confirmed the strong discriminative performancen between FTD and controls (AUCs > 0.96) and between FTD and AD (AUCs > 0.88) across three validation cohorts, including one with autopsy confirmation (AUCs > 0.90). Validation in genetic FTD (including C9orf72 , GRN , and MAPT mutation carriers) revealed high accuracy of the FTD diagnostic panel in identifying both the presymptomatic (AUCs > 0.95) and symptomatic (AUC: 1) stages. Six proteins were differentially regulated between FTLD-TDP and FTLD-Tau. However, a reproducible classification model could not be generated (AUC: 0.80). Conclusions: Overall, this study introduces novel FTD-specific biomarker panels with potential use in diagnostic settings.
- Open Access
- 11 August 2025
- Published by Springer Nature in Metabolomics
- Vol. 21 (5) , 105
Estimating metabolite networks subject to dietary preferences and lifestyle
Introduction: The metabolome is an intermediate between DNA variation and clinical phenotypes. Metabolomics have been widely used in biomedical studies for reflecting physiological changes in response to variation coming from various sources, such as diet, environment, time, and lifestyle. While lifestyle factors contribute a considerable part of the metabolic variation, current human studies lack information estimating lifestyle, mainly because it is not strictly defined. Objective: In this work, metabolite concentrations are measured at two time points (2007 and 2014). Additionally, SNP data together with self-reports on dietary behavior. By having measurements over time, as well as all main sources of metabolic variation (diet, genetics), both time-effects and lifestyle-effects can be estimated. Since lifestyle and time effects can be estimated under this setting, we are interested in identifying metabolites sharing similar relationships to diet and lifestyle, using network analysis. Methods: The correlation between repeated measurements is modeled using a random intercepts linear mixed model, with dietary preferences, genetics, and time as fixed effects. The random intercepts can be defined as the lifestyle, and represent the part of the metabolic variation which is not due to diet, genetics, and time and is subject-specific. The part of every metabolite relevant to diet and lifestyle instead of the original values is used as input values to network estimation methods. Conclusions: This work demonstrates how correcting for several sources of metabolic variation, allows us to look for residual variation and build networks with meaningful metabolite groups sharing similar association to diet and lifestyle.
- 11 August 2025
- Published by Springer Nature in Metabolomics
- Vol. 21 (5) , 105
Open AccessEstimating metabolite networks subject to dietary preferences and lifestyle
Introduction: The metabolome is an intermediate between DNA variation and clinical phenotypes. Metabolomics have been widely used in biomedical studies for reflecting physiological changes in response to variation coming from various sources, such as diet, environment, time, and lifestyle. While lifestyle factors contribute a considerable part of the metabolic variation, current human studies lack information estimating lifestyle, mainly because it is not strictly defined. Objective: In this work, metabolite concentrations are measured at two time points (2007 and 2014). Additionally, SNP data together with self-reports on dietary behavior. By having measurements over time, as well as all main sources of metabolic variation (diet, genetics), both time-effects and lifestyle-effects can be estimated. Since lifestyle and time effects can be estimated under this setting, we are interested in identifying metabolites sharing similar relationships to diet and lifestyle, using network analysis. Methods: The correlation between repeated measurements is modeled using a random intercepts linear mixed model, with dietary preferences, genetics, and time as fixed effects. The random intercepts can be defined as the lifestyle, and represent the part of the metabolic variation which is not due to diet, genetics, and time and is subject-specific. The part of every metabolite relevant to diet and lifestyle instead of the original values is used as input values to network estimation methods. Conclusions: This work demonstrates how correcting for several sources of metabolic variation, allows us to look for residual variation and build networks with meaningful metabolite groups sharing similar association to diet and lifestyle.
- Open Access
DNA methylation networks during pig fetal development: a joint fused ridge estimation approach
Although an organism’s genetic information is predominantly identical among most of its cell types, the epigenome regulates the expression of the genome in a cell type- and context-dependent manner. In mammals, DNA methylation in regulatory regions, such as promoters, primarily regulates gene expression by inducing transcriptional inactivation. With genome-wide approaches came the realization that DNA methylation patterns underlying mammalian development are considerably more dynamic than previously recognized. This realization highlights the need for methodological approaches capable of capturing this phenomenon. In this study, we investigated the feasibility of modeling DNA methylation networks by jointly estimating regularized precision matrices from time- and tissue-specific omics data derived from the pig genome. For that, we utilized RNA-Seq and RRBS data that span seven pig tissues at three developmental stages: early organogenesis, late organogenesis, and newborn. Our analysis focused on 61, 48, and 74 genes — differentially expressed across developmental stages and CpG-methylated in promoter regions, from endoderm-, mesoderm-, and ectoderm-derived tissues, respectively. Using a joint fused ridge approach, we were able to borrow information across tissues and time points, enabling more robust network inference. This analytical framework advances exploratory methods for studying organism development using pig as a model species. Our results highlight the importance of fetal-maternal immunity and the circulatory system in early development, and shed light on dynamic interactions across tissues, organ systems, and germ layers. We anticipate that this flexible framework can be extended to other omics data and species, facilitating future research. Author summary: This study explores how gene activity is controlled during pig fetal development using networks. The epigenome can turn genes on and off, depending on, e.g. the cell’s function and the organism’s growth stage. This process includes chemical changes to DNA, such as methylation. We focused on DNA methylation patterns during pig fetal development by analyzing samples from several organs at three stages of fetal growth. We combined gene activity data with DNA methylation data by using a network-based method that allows us to visualize and study how genes interact across different tissues and developmental stages. We discovered that the immune and circulatory systems are critical during early development. We have also observed complex interactions across tissues and organ systems. Our study provides a flexible toolbox with clear, step-by-step explanations. This analytical framework can be used to explore developmental patterns at the genomic scale in other species and can be adapted to different types of biological data.
Open AccessDNA methylation networks during pig fetal development: a joint fused ridge estimation approach
Although an organism’s genetic information is predominantly identical among most of its cell types, the epigenome regulates the expression of the genome in a cell type- and context-dependent manner. In mammals, DNA methylation in regulatory regions, such as promoters, primarily regulates gene expression by inducing transcriptional inactivation. With genome-wide approaches came the realization that DNA methylation patterns underlying mammalian development are considerably more dynamic than previously recognized. This realization highlights the need for methodological approaches capable of capturing this phenomenon. In this study, we investigated the feasibility of modeling DNA methylation networks by jointly estimating regularized precision matrices from time- and tissue-specific omics data derived from the pig genome. For that, we utilized RNA-Seq and RRBS data that span seven pig tissues at three developmental stages: early organogenesis, late organogenesis, and newborn. Our analysis focused on 61, 48, and 74 genes — differentially expressed across developmental stages and CpG-methylated in promoter regions, from endoderm-, mesoderm-, and ectoderm-derived tissues, respectively. Using a joint fused ridge approach, we were able to borrow information across tissues and time points, enabling more robust network inference. This analytical framework advances exploratory methods for studying organism development using pig as a model species. Our results highlight the importance of fetal-maternal immunity and the circulatory system in early development, and shed light on dynamic interactions across tissues, organ systems, and germ layers. We anticipate that this flexible framework can be extended to other omics data and species, facilitating future research. Author summary: This study explores how gene activity is controlled during pig fetal development using networks. The epigenome can turn genes on and off, depending on, e.g. the cell’s function and the organism’s growth stage. This process includes chemical changes to DNA, such as methylation. We focused on DNA methylation patterns during pig fetal development by analyzing samples from several organs at three stages of fetal growth. We combined gene activity data with DNA methylation data by using a network-based method that allows us to visualize and study how genes interact across different tissues and developmental stages. We discovered that the immune and circulatory systems are critical during early development. We have also observed complex interactions across tissues and organ systems. Our study provides a flexible toolbox with clear, step-by-step explanations. This analytical framework can be used to explore developmental patterns at the genomic scale in other species and can be adapted to different types of biological data.
- Open Access
Large-scale CSF proteome profiling identifies biomarkers for accurate diagnosis of Frontotemporal Dementia
Diagnosis of Frontotemporal dementia (FTD) and the specific underlying neuropathologies (frontotemporal lobar degeneration; FTLD-Tau and FTLD-TDP) is challenging, and thus fluid biomarkers are needed to improve diagnostic accuracy. We used proximity extension assays to analyze 665 proteins in cerebrospinal fluid (CSF) samples from a multicenter cohort including patients with FTD (n = 189), Alzheimer’s Disease dementia (AD; n = 232), and cognitively unimpaired individuals (n = 196). In a subset, FTLD neuropathology was determined based on phenotype or genotype (FTLD-Tau = 87 and FTLD-TDP = 68). Forty three proteins were differentially regulated in FTD compared to controls and AD, reflecting axon development, regulation of synapse assembly, and cell-cell adhesion mediator activity pathways. Classification analysis identified a 14- and 13-CSF protein panel that discriminated FTD from controls (AUC: 0.96) or AD (AUC: 0.91). Custom multiplex panels confirmed the highly accurate discrimination between FTD and controls (AUCs > 0.96) or AD (AUCs > 0.88) in three validation cohorts, including one with autopsy confirmation (AUCs > 0.90). Six proteins were differentially regulated between FTLD-TDP and FTLD-Tau, but no reproducible classification model could be generated (AUC: 0.80). Overall, this study introduces novel FTD-specific biomarker panels with potential use in diagnostic setting.
Open AccessLarge-scale CSF proteome profiling identifies biomarkers for accurate diagnosis of Frontotemporal Dementia
Diagnosis of Frontotemporal dementia (FTD) and the specific underlying neuropathologies (frontotemporal lobar degeneration; FTLD-Tau and FTLD-TDP) is challenging, and thus fluid biomarkers are needed to improve diagnostic accuracy. We used proximity extension assays to analyze 665 proteins in cerebrospinal fluid (CSF) samples from a multicenter cohort including patients with FTD (n = 189), Alzheimer’s Disease dementia (AD; n = 232), and cognitively unimpaired individuals (n = 196). In a subset, FTLD neuropathology was determined based on phenotype or genotype (FTLD-Tau = 87 and FTLD-TDP = 68). Forty three proteins were differentially regulated in FTD compared to controls and AD, reflecting axon development, regulation of synapse assembly, and cell-cell adhesion mediator activity pathways. Classification analysis identified a 14- and 13-CSF protein panel that discriminated FTD from controls (AUC: 0.96) or AD (AUC: 0.91). Custom multiplex panels confirmed the highly accurate discrimination between FTD and controls (AUCs > 0.96) or AD (AUCs > 0.88) in three validation cohorts, including one with autopsy confirmation (AUCs > 0.90). Six proteins were differentially regulated between FTLD-TDP and FTLD-Tau, but no reproducible classification model could be generated (AUC: 0.80). Overall, this study introduces novel FTD-specific biomarker panels with potential use in diagnostic setting.
- Open Access
- 30 May 2024
- Published by Springer Nature in BMC Bioinformatics
- Vol. 25 (1) , 1-17
A guided network estimation approach using multi-omic information
In systems biology, an organism is viewed as a system of interconnected molecular entities. To understand the functioning of organisms it is essential to integrate information about the variations in the concentrations of those molecular entities. This information can be structured as a set of networks with interconnections and with some hierarchical relations between them. Few methods exist for the reconstruction of integrative networks. In this work, we propose an integrative network reconstruction method in which the network organization for a particular type of omics data is guided by the network structure of a related type of omics data upstream in the omic cascade. The structure of these guiding data can be either already known or be estimated from the guiding data themselves. The method consists of three steps. First a network structure for the guiding data should be provided. Next, responses in the target set are regressed on the full set of predictors in the guiding data with a Lasso penalty to reduce the number of predictors and an L2 penalty on the differences between coefficients for predictors that share edges in the network for the guiding data. Finally, a network is reconstructed on the fitted target responses as functions of the predictors in the guiding data. This way we condition the target network on the network of the guiding data. We illustrate our approach on two examples in Arabidopsis. The method detects groups of metabolites that have a similar genetic or transcriptomic basis.
- 30 May 2024
- Published by Springer Nature in BMC Bioinformatics
- Vol. 25 (1) , 1-17
Open AccessA guided network estimation approach using multi-omic information
In systems biology, an organism is viewed as a system of interconnected molecular entities. To understand the functioning of organisms it is essential to integrate information about the variations in the concentrations of those molecular entities. This information can be structured as a set of networks with interconnections and with some hierarchical relations between them. Few methods exist for the reconstruction of integrative networks. In this work, we propose an integrative network reconstruction method in which the network organization for a particular type of omics data is guided by the network structure of a related type of omics data upstream in the omic cascade. The structure of these guiding data can be either already known or be estimated from the guiding data themselves. The method consists of three steps. First a network structure for the guiding data should be provided. Next, responses in the target set are regressed on the full set of predictors in the guiding data with a Lasso penalty to reduce the number of predictors and an L2 penalty on the differences between coefficients for predictors that share edges in the network for the guiding data. Finally, a network is reconstructed on the fitted target responses as functions of the predictors in the guiding data. This way we condition the target network on the network of the guiding data. We illustrate our approach on two examples in Arabidopsis. The method detects groups of metabolites that have a similar genetic or transcriptomic basis.
- Open Access
- 22 October 2022
- Published by Springer Nature in Acta Neuropathologica Communications
- Vol. 10 (1) , 1-11
Plasma proteome profiling identifies changes associated to AD but not to FTD
Background: Frontotemporal dementia (FTD) is caused by frontotemporal lobar degeneration (FTLD), characterized mainly by inclusions of Tau (FTLD-Tau) or TAR DNA binding43 (FTLD-TDP) proteins. Plasma biomarkers are strongly needed for specific diagnosis and potential treatment monitoring of FTD. We aimed to identify specific FTD plasma biomarker profiles discriminating FTD from AD and controls, and between FTD pathological subtypes. In addition, we compared plasma results with results in post-mortem frontal cortex of FTD cases to understand the underlying process. Methods: Plasma proteins (n = 1303) from pathologically and/or genetically confirmed FTD patients (n = 56; FTLD-Tau n = 16; age = 58.2 ± 6.2; 44% female, FTLD-TDP n = 40; age = 59.8 ± 7.9; 45% female), AD patients (n = 57; age = 65.5 ± 8.0; 39% female), and non-demented controls (n = 148; 61.3 ± 7.9; 41% female) were measured using an aptamer-based proteomic technology (SomaScan). In addition, exploratory analysis in post-mortem frontal brain cortex of FTD (n = 10; FTLD-Tau n = 5; age = 56.2 ± 6.9, 60% female, and FTLD-TDP n = 5; age = 64.0 ± 7.7, 60% female) and non-demented controls (n = 4; age = 61.3 ± 8.1; 75% female) were also performed. Differentially regulated plasma and tissue proteins were identified by global testing adjusting for demographic variables and multiple testing. Logistic lasso regression was used to identify plasma protein panels discriminating FTD from non-demented controls and AD, or FTLD-Tau from FTLD-TDP. Performance of the discriminatory plasma protein panels was based on predictions obtained from bootstrapping with 1000 resampled analysis. Results: Overall plasma protein expression profiles differed between FTD, AD and controls (6 proteins; p = 0.005), but none of the plasma proteins was specifically associated to FTD. The overall tissue protein expression profile differed between FTD and controls (7-proteins; p = 0.003). There was no difference in overall plasma or tissue expression profile between FTD subtypes. Regression analysis revealed a panel of 12-plasma proteins discriminating FTD from AD with high accuracy (AUC: 0.99). No plasma protein panels discriminating FTD from controls or FTD pathological subtypes were identified. Conclusions: We identified a promising plasma protein panel as a minimally-invasive tool to aid in the differential diagnosis of FTD from AD, which was primarily associated to AD pathophysiology. The lack of plasma profiles specifically associated to FTD or its pathological subtypes might be explained by FTD heterogeneity, calling for FTD studies using large and well-characterize cohorts.
- 22 October 2022
- Published by Springer Nature in Acta Neuropathologica Communications
- Vol. 10 (1) , 1-11
Open AccessPlasma proteome profiling identifies changes associated to AD but not to FTD
Background: Frontotemporal dementia (FTD) is caused by frontotemporal lobar degeneration (FTLD), characterized mainly by inclusions of Tau (FTLD-Tau) or TAR DNA binding43 (FTLD-TDP) proteins. Plasma biomarkers are strongly needed for specific diagnosis and potential treatment monitoring of FTD. We aimed to identify specific FTD plasma biomarker profiles discriminating FTD from AD and controls, and between FTD pathological subtypes. In addition, we compared plasma results with results in post-mortem frontal cortex of FTD cases to understand the underlying process. Methods: Plasma proteins (n = 1303) from pathologically and/or genetically confirmed FTD patients (n = 56; FTLD-Tau n = 16; age = 58.2 ± 6.2; 44% female, FTLD-TDP n = 40; age = 59.8 ± 7.9; 45% female), AD patients (n = 57; age = 65.5 ± 8.0; 39% female), and non-demented controls (n = 148; 61.3 ± 7.9; 41% female) were measured using an aptamer-based proteomic technology (SomaScan). In addition, exploratory analysis in post-mortem frontal brain cortex of FTD (n = 10; FTLD-Tau n = 5; age = 56.2 ± 6.9, 60% female, and FTLD-TDP n = 5; age = 64.0 ± 7.7, 60% female) and non-demented controls (n = 4; age = 61.3 ± 8.1; 75% female) were also performed. Differentially regulated plasma and tissue proteins were identified by global testing adjusting for demographic variables and multiple testing. Logistic lasso regression was used to identify plasma protein panels discriminating FTD from non-demented controls and AD, or FTLD-Tau from FTLD-TDP. Performance of the discriminatory plasma protein panels was based on predictions obtained from bootstrapping with 1000 resampled analysis. Results: Overall plasma protein expression profiles differed between FTD, AD and controls (6 proteins; p = 0.005), but none of the plasma proteins was specifically associated to FTD. The overall tissue protein expression profile differed between FTD and controls (7-proteins; p = 0.003). There was no difference in overall plasma or tissue expression profile between FTD subtypes. Regression analysis revealed a panel of 12-plasma proteins discriminating FTD from AD with high accuracy (AUC: 0.99). No plasma protein panels discriminating FTD from controls or FTD pathological subtypes were identified. Conclusions: We identified a promising plasma protein panel as a minimally-invasive tool to aid in the differential diagnosis of FTD from AD, which was primarily associated to AD pathophysiology. The lack of plasma profiles specifically associated to FTD or its pathological subtypes might be explained by FTD heterogeneity, calling for FTD studies using large and well-characterize cohorts.
- Open Access
- 28 September 2022
- Published by MDPI AG in Current Oncology
- Vol. 29 (10) , 7109-7121
Psychoneurological Symptoms and Biomarkers of Stress and Inflammation in Newly Diagnosed Head and Neck Cancer Patients: A Network Analysis
Psychoneurological symptoms are commonly reported by newly diagnosed head and neck cancer (HNC) patients, yet there is limited research on the associations of these symptoms with biomarkers of stress and inflammation. In this article, pre-treatment data of a multi-center cohort of HNC patients were analyzed using a network analysis to examine connections between symptoms (poor sleep quality, anxiety, depression, fatigue, and oral pain), biomarkers of stress (diurnal cortisol slope), inflammation markers (c-reactive protein [CRP], interleukin [IL]-6, IL-10, and tumor necrosis factor alpha [TNF-α]), and covariates (age and body mass index [BMI]). Three centrality indices were calculated: degree (number of connections), closeness (proximity of a variable to other variables), and betweenness (based on the number of times a variable is located on the shortest path between any pair of other variables). In a sample of 264 patients, poor sleep quality and fatigue had the highest degree index; fatigue and CRP had the highest closeness index; and IL-6 had the highest betweenness index. The model yielded two clusters: a symptoms—cortisol slope—CRP cluster and a IL-6—IL-10—TNF-α—age—BMI cluster. Both clusters were connected most prominently via IL-6. Our findings provide evidence that poor sleep quality, fatigue, CRP, and IL-6 play an important role in the interconnections between psychoneurological symptoms and biomarkers of stress and inflammation in newly diagnosed HNC patients.
- 28 September 2022
- Published by MDPI AG in Current Oncology
- Vol. 29 (10) , 7109-7121
Open AccessPsychoneurological Symptoms and Biomarkers of Stress and Inflammation in Newly Diagnosed Head and Neck Cancer Patients: A Network Analysis
Psychoneurological symptoms are commonly reported by newly diagnosed head and neck cancer (HNC) patients, yet there is limited research on the associations of these symptoms with biomarkers of stress and inflammation. In this article, pre-treatment data of a multi-center cohort of HNC patients were analyzed using a network analysis to examine connections between symptoms (poor sleep quality, anxiety, depression, fatigue, and oral pain), biomarkers of stress (diurnal cortisol slope), inflammation markers (c-reactive protein [CRP], interleukin [IL]-6, IL-10, and tumor necrosis factor alpha [TNF-α]), and covariates (age and body mass index [BMI]). Three centrality indices were calculated: degree (number of connections), closeness (proximity of a variable to other variables), and betweenness (based on the number of times a variable is located on the shortest path between any pair of other variables). In a sample of 264 patients, poor sleep quality and fatigue had the highest degree index; fatigue and CRP had the highest closeness index; and IL-6 had the highest betweenness index. The model yielded two clusters: a symptoms—cortisol slope—CRP cluster and a IL-6—IL-10—TNF-α—age—BMI cluster. Both clusters were connected most prominently via IL-6. Our findings provide evidence that poor sleep quality, fatigue, CRP, and IL-6 play an important role in the interconnections between psychoneurological symptoms and biomarkers of stress and inflammation in newly diagnosed HNC patients.
- Open Access
- 22 June 2022
- Published by Wiley in Biometrical Journal
- Vol. 64 (7) , 1289-1306
Semi‐supervised empirical Bayes group‐regularized factor regression
The features in a high‐dimensional biomedical prediction problem are often well described by low‐dimensional latent variables (or factors). We use this to include unlabeled features and additional information on the features when building a prediction model. Such additional feature information is often available in biomedical applications. Examples are annotation of genes, metabolites, or p‐values from a previous study. We employ a Bayesian factor regression model that jointly models the features and the outcome using Gaussian latent variables. We fit the model using a computationally efficient variational Bayes method, which scales to high dimensions. We use the extra information to set up a prior model for the features in terms of hyperparameters, which are then estimated through empirical Bayes. The method is demonstrated in simulations and two applications. One application considers influenza vaccine efficacy prediction based on microarray data. The second application predicts oral cancer metastasis from RNAseq data.
- 22 June 2022
- Published by Wiley in Biometrical Journal
- Vol. 64 (7) , 1289-1306
Open AccessSemi‐supervised empirical Bayes group‐regularized factor regression
The features in a high‐dimensional biomedical prediction problem are often well described by low‐dimensional latent variables (or factors). We use this to include unlabeled features and additional information on the features when building a prediction model. Such additional feature information is often available in biomedical applications. Examples are annotation of genes, metabolites, or p‐values from a previous study. We employ a Bayesian factor regression model that jointly models the features and the outcome using Gaussian latent variables. We fit the model using a computationally efficient variational Bayes method, which scales to high dimensions. We use the extra information to set up a prior model for the features in terms of hyperparameters, which are then estimated through empirical Bayes. The method is demonstrated in simulations and two applications. One application considers influenza vaccine efficacy prediction based on microarray data. The second application predicts oral cancer metastasis from RNAseq data.
- Open Access
- 8 April 2022
- Published by Springer Nature in Euphytica
- Vol. 218 (5) , 1-14
psBLUP: incorporating marker proximity for improving genomic prediction accuracy
Genomic selection entails the estimation of phenotypic traits of interest for plants without phenotype based on the association between single-nucleotide polymorphisms (SNPs) and phenotypic traits for plants with phenotype. Typically, the number of SNPs far exceeds the number of samples (high-dimensionality) and, therefore, usage of regularization methods is common. The most common approach to estimate marker-trait associations uses the genomic best linear unbiased predictor (GBLUP) method, where a mixed model is fitted to the data. GBLUP has also been alternatively parameterized as a ridge regression model (RRBLUP). GBLUP/RRBLUP is based on the assumption of independence between predictor variables. However, it is to be expected that variables will be associated due to their genetic proximity. Here, we propose a regularized linear model (namely psBLUP: proximity smoothed BLUP) that explicitly models the dependence between predictor effects. We show that psBLUP can improve accuracy compared to the standard methods on both Arabidopsis thaliana data and Barley data.
- 8 April 2022
- Published by Springer Nature in Euphytica
- Vol. 218 (5) , 1-14
Open AccesspsBLUP: incorporating marker proximity for improving genomic prediction accuracy
Genomic selection entails the estimation of phenotypic traits of interest for plants without phenotype based on the association between single-nucleotide polymorphisms (SNPs) and phenotypic traits for plants with phenotype. Typically, the number of SNPs far exceeds the number of samples (high-dimensionality) and, therefore, usage of regularization methods is common. The most common approach to estimate marker-trait associations uses the genomic best linear unbiased predictor (GBLUP) method, where a mixed model is fitted to the data. GBLUP has also been alternatively parameterized as a ridge regression model (RRBLUP). GBLUP/RRBLUP is based on the assumption of independence between predictor variables. However, it is to be expected that variables will be associated due to their genetic proximity. Here, we propose a regularized linear model (namely psBLUP: proximity smoothed BLUP) that explicitly models the dependence between predictor effects. We show that psBLUP can improve accuracy compared to the standard methods on both Arabidopsis thaliana data and Barley data.