Publications
-
-
Parekh, Pravesh; Parker, Nadine; Pecheva, Diliana; Frei, Evgeniia; Vaudel, Marc & Smith, Diana M.
[Show all 26 contributors for this article]
(2026).
FEMA-Long: Modeling unstructured covariances for discovery of time-dependent effects in large-scale longitudinal datasets.
PLoS Genetics.
ISSN 1553-7390.
22(6).
doi:
10.1371/journal.pgen.1012184.
Full text in Research Archive
Show summary
While linear mixed-effects (LME) models are common for analyzing longitudinal data, most users rely on random intercepts or simple stationary covariance, due to unavailability of computationally tractable solutions. Here, we extend the Fast and Efficient Mixed-Effects Algorithm (FEMA) and present FEMA-Long, a computationally tractable approach to flexibly modeling longitudinal covariance suitable for high-dimensional data. FEMA-Long can: i) model unstructured covariance, ii) model covariates as smooth functions using splines, iii) discover time-dependent effects of covariates with spline interactions, and iv) use these flexible longitudinal modeling strategies to perform longitudinal genome-wide association studies and discover time-dependent genetic effects, in a computationally scalable manner, suitable for high-dimensional data. Through extensive simulations, we show that estimates from FEMA-Long are accurate, while being up to several thousand times faster and with minimal carbon footprint. To show the utility of FEMA-Long for discovering novel biological signal, using data from the Norwegian Mother, Father and Child Cohort Study (MoBa), we performed a longitudinal genome-wide association study with non-linear SNP-by-time interaction on length, weight, and BMI of 68,273 infants with up to six measurements in the first year of life. We found dynamic patterns of random effects including time-varying heritability and genetic correlations, as well as several genetic variants showing time-dependent effects, highlighting the applicability of FEMA-Long to enable novel discoveries.
-
Frei, Evgeniia; Frei, Oleksandr; Hagen, Espen; Shadrin, Alexey; Bakken, Nora Refsum & Birkenæs, Viktoria
[Show all 9 contributors for this article]
(2026).
Prediction of adolescent internalizing disorder risk: Evidence from the Norwegian mother, father, and child cohort study.
Journal of Affective Disorders.
ISSN 0165-0327.
404,
p. 1–9.
doi:
10.1016/j.jad.2026.121481.
Full text in Research Archive
Show summary
Background
Internalizing disorders are among the most common psychiatric conditions in adolescence, often associated with long-term adverse outcomes. Early identification of at-risk youth is important for effective intervention, though it remains challenging due to the multifactorial nature of risk. Machine learning (ML) offers opportunities to integrate multiple data sources and improve risk prediction for internalizing disorders.
Methods
We used data from 13,743 adolescents (mean age 14.45 years; 52.7% female) participating in the Norwegian Mother, Father and Child Cohort Study (MoBa), linked to national health registries. Logistic regression with elastic net regularization was applied to predict the risk of an internalizing disorder (mood, anxiety or stress-related) occurring within one to five years after assessment. Nested models of increasing complexity incorporated sociodemographic, clinical, lifestyle, mental health, psychosocial, and genetic predictors. Model performance was evaluated in a hold-out test set. Simplified models combining three questionnaire scales were also evaluated.
Results
Test-set performance increased with model complexity, reaching area under the receiver operating characteristic curve (AUC) of 0.731 for the full model. Mental health self-reported symptoms and psychosocial predictors contributed most to the discrimination. Simplified models using three questionnaire scales, alongside age and sex, achieved AUCs up to 0.718 and effectively stratified adolescents into high- and low-risk groups (OR80/20 ranged 6.11–9.35).
Conclusion
Multimodal ML models integrating registry information, mental health symptoms, psychosocial factors, and genetic data demonstrated moderate predictive performance. Simplified models with three questionnaire items reached comparable performance, highlighting their potential utility in the early identification of adolescents at elevated internalizing disorder risk.
-
-
-
Birkenæs, Viktoria; Parekh, Pravesh; Hegemann, Laura Elizabeth; Bakken, Nora Refsum; Frei, Evgeniia & Jaholkowski, Piotr Pawel
[Show all 13 contributors for this article]
(2025).
Youth psychotic experiences: psychometric evaluation and diagnostic associations of the CAPE-16 in adolescents from the Norwegian Mother, Mather and Child Cohort.
Journal of Child Psychology and Psychiatry.
ISSN 0021-9630.
67(1),
p. 27–40.
doi:
10.1111/jcpp.70007.
Full text in Research Archive
Show summary
Background: Adolescent self-reported psychotic experiences are associated with mental illness and could help guide prevention strategies. Youth report substantially more experiences than adults. However, with large societal changes like the digital revolution and COVID-19 pandemic, existing questionnaires may no longer accurately capture youth experiences. We aimed to determine the ability of the CAPE-16 questionnaire in capturing psychotic experiences across contexts (biological sex and COVID-19 response) and generations, thereby validating important psychometric aspects of the tool in modern adolescents.
Methods: We used data from the Norwegian Mother, Father and Child Study (MoBa), a population-based pregnancy cohort. Adolescents responded to the CAPE-16 questionnaire (n = 18,835). For a comparison between age groups, we included adult men from the parent generation who responded to the CAPE-9 (n = 28,793). We investigated the psychometric properties of CAPE-16 through confirmatory factor analyses, measurement invariance testing across biological sex, response before/during the COVID-19 pandemic, and generations (adolescents and fathers), and examined subscale and item-level associations with subsequent registry-based psychiatric diagnoses (average time between CAPE and last registry update: 3.68 ± 1.34 years).
Results: Out of 18,835 adolescents, 33.2% reported lifetime psychotic experiences. We confirmed a three-factor structure (paranoia, bizarre thoughts, and hallucinations) and good subscale reliability (ω = .86 and .90). CAPE-16 scores were stable across biological sex and pandemic status. CAPE-9 response patterns were non-invariant across adolescents and adult men, with an item related to digital technology particularly prone to bias. CAPE-16 subscales were associated with subsequent psychiatric diagnoses, especially psychotic disorders.
Conclusions: CAPE-16 is a reliable measure of psychotic experiences across sex and a major societal stressor in adolescents. More frequent and distressing experiences increase the risk of subsequent psychiatric diagnoses. Different response patterns between adults and adolescents for items related to digital technology suggest differences in interpretation. Hence, certain items may benefit from revisions.
-
Jaholkowski, Piotr Pawel; Bahrami, Shahram; Fominykh, Vera; Hindley, Guy; Tesfaye, Markos & Parekh, Pravesh
[Show all 23 contributors for this article]
(2024).
Charting the shared genetic architecture of Alzheimer's disease, cognition, and educational attainment, and associations with brain development.
Neurobiology of Disease.
ISSN 0969-9961.
203.
doi:
10.1016/j.nbd.2024.106750.
Full text in Research Archive
Show summary
The observation that the risk of developing Alzheimer's disease is reduced in individuals with high premorbid cognitive functioning, higher educational attainment, and occupational status has led to the ‘cognitive reserve’ hypothesis. This hypothesis suggests that individuals with greater cognitive reserve can tolerate a more significant burden of neuropathological changes before the onset of cognitive decline. The underpinnings of cognitive reserve remain poorly understood, although a shared genetic basis between measures of cognitive reserve and Alzheimer's disease has been suggested. Using the largest samples to date and novel statistical tools, we aimed to investigate shared genetic variants between Alzheimer's disease, and measures of cognitive reserve; cognition and educational attainment to identify molecular and neurobiological foundations. We applied the causal mixture model (MiXeR) to estimate the number of trait-influencing variants shared between Alzheimer's disease, cognition, and educational attainment, and condFDR/conjFDR to identify shared loci. To provide biological insights loci were functionally characterized. Subsequently, we constructed a Structural Equation Model (SEM) to determine if the polygenic foundation of cognition has a direct impact on Alzheimer's disease risk, or if its effect is mediated through established risk factors for the disease, using a case-control sample from the UK Biobank. Univariate MiXeR analysis (after excluding chromosome 19) revealed that Alzheimer's disease was substantially less polygenic (450 trait-influencing variants) compared to cognition (11,100 trait-influencing variants), and educational attainment (12,700 trait-influencing variants). Bivariate MiXeR analysis estimated that Alzheimer's disease shared approximately 70 % of trait-influencing variants with cognition, and approximately 40 % with educational attainment, with mixed effect directions. Using condFDR analysis, we identified 18 loci jointly associated with Alzheimer's disease and cognition and 6 loci jointly associated with Alzheimer's disease and educational attainment. Genes mapped to shared loci were associated with neurodevelopment, expressed in early life, and implicated the dendritic tree and phosphatidylinositol phosphate binding mechanisms. Spatiotemporal gene expression analysis of the identified genes showed that mapped genes were highly expressed during the mid-fetal period, further suggesting early neurodevelopmental stages as critical periods for establishing cognitive reserve which affect the risk of Alzheimer's disease in old age. Furthermore, our SEM analysis showed that genetic variants influencing cognition had a direct effect on the risk of developing Alzheimer's disease, providing evidence in support of the neurodevelopmental hypothesis of the disease.
-
Bakken, Nora Refsum; Parker, Nadine; Hannigan, Laurie John; Hagen, Espen; Parekh, Pravesh & Shadrin, Alexey
[Show all 15 contributors for this article]
(2024).
Childhood trajectories of emotional and behavioral difficulties are related to polygenic liability for mood and anxiety disorders.
Journal of Child Psychology and Psychiatry.
ISSN 0021-9630.
66(3),
p. 350–365.
doi:
10.1111/jcpp.14063.
Full text in Research Archive
Show summary
Background Symptoms related to mood and anxiety disorders (emotional disorders) often present in childhood and adolescence. Some of the genetic liability for mental disorders, and emotional and behavioral difficulties seems to be shared. Yet, it is unclear how genetic liability for emotional disorders and related traits influence trajectories of childhood behavioral and emotional difficulties, and if specific developmental patterns are associated with higher genetic liability for these disorders. Methods This study uses data from a genotyped sample of children (n = 54,839) from the Norwegian Mother, Father, and Child Cohort Study (MoBa). We use latent growth models (1.5–5 years) and latent profile analyses (1.5–8 years) to quantify childhood trajectories and profiles of emotional and behavioral difficulties and diagnoses. We examine associations between these trajectories and profiles with polygenic scores for bipolar disorder (PGSBD), anxiety (PGSANX), depression (PGSDEP), and neuroticism (PGSNEUR). Results Associations between PGSDEP, PGSANX, and PGSNEUR, and emotional and behavioral difficulties in childhood were more persistent than age-specific across early childhood (1.5–5 years). Higher PGSANX and PGSDEP were associated with steeper increases in behavioral difficulties across early childhood. Latent profile analyses identified five profiles with different associations with emotional disorder diagnosis. All PGS were associated with the probability of classification into profiles characterized by some form of difficulties (vs. a normative reference profile), but only PGSBD was uniquely associated with a single developmental profile. Conclusions Genetic risk for mood disorders and related traits contribute to both a higher baseline level of, and a more rapid increase in, emotional and behavioral difficulties across early and middle childhood, with some indications for disorder-specific profiles. Our findings may inform research on developmental pathways to emotional disorders and the improvement of initiatives for early identification and targeted intervention.
-
-
-
Birkenæs, Viktoria; Bakken, Nora Refsum; Frei, Evgeniia; Jaholkowski, Piotr Pawel; Smeland, Olav Bjerkehagen & Woldeyohannes, Markos Tesfaye
[Show all 15 contributors for this article]
(2023).
Psychometric Properties and Diagnostic Associations of the Short-Form Community Assessment of Psychic Experiences in a Population-Based Sample of 29 021 Adult Men.
Schizophrenia Bulletin.
ISSN 0586-7614.
49(5),
p. 1229–1238.
doi:
10.1093/schbul/sbad074.
Full text in Research Archive
Show summary
Abstract Background and Hypothesis Around 5%–7% of the adult population are estimated to have lifetime psychotic experiences (PEs), which are associated with psychosis risk. PEs assessed with Community Assessment of Psychic Experiences (CAPE) are associated with psychosis but also non-psychotic disorders, which could be partly explained by CAPE indirectly capturing emotional symptoms. We investigated the psychometric properties of a shorter version, CAPE-9, and whether CAPE-9 scores are associated with lifetime psychotic or non-psychotic mental disorders after controlling for current anxiety and depressive symptoms. Design CAPE-9 questionnaire data were obtained from 29 021 men (42.4 ± 5.6 yrs.) from the Norwegian Mother, Father, and Child Cohort Study. We investigated CAPE-9 reliability and factor structure. Logistic regression was used to test effects of current anxiety and depressive symptoms (SCL-12) on associations between CAPE-9 scores and psychiatric diagnoses. Results CAPE-9 fit a previously reported 3-factor structure and showed good reliability. Twenty-six percent reported at least one lifetime PE. CAPE-9 scores were significantly associated with most psychiatric disorders (schizophrenia, depression, bipolar disorder, substance abuse, anxiety, trauma-related disorders, and ADHD). After controlling for concurrent emotional symptoms, only associations with schizophrenia (OR = 1.29; 95% CI = 1.18–1.38) and trauma-related disorders (OR = 1.09; CI = 1.02–1.15) remained significant. Conclusions CAPE-9 showed good psychometric properties in this large population-based adult male sample, and PEs were more clearly associated with psychotic disorders after controlling for current emotional symptoms. These results support the use of the short CAPE-9 as a cost-effective tool for informing public health initiatives and advancing our understanding of the dimensionality of psychosis.
-
-
Assis, Denis Reis de; Szabo, Attila; Osete, Jordi Requena; Puppo, Francesca; O'Connell, Kevin S. & Akkouh, Ibrahim
[Show all 10 contributors for this article]
(2021).
Using iPSC Models to Understand the Role of Estrogen in Neuron–Glia Interactions in Schizophrenia and Bipolar Disorder.
Cells.
10(2).
doi:
10.3390/cells10020209.
Full text in Research Archive
Show summary
Schizophrenia (SCZ) and bipolar disorder (BIP) are severe mental disorders with a considerable disease burden worldwide due to early age of onset, chronicity, and lack of efficient treatments or prevention strategies. Whilst our current knowledge is that SCZ and BIP are highly heritable and share common pathophysiological mechanisms associated with cellular signaling, neurotransmission, energy metabolism, and neuroinflammation, the development of novel therapies has been hampered by the unavailability of appropriate models to identify novel targetable pathomechanisms. Recent data suggest that neuron–glia interactions are disturbed in SCZ and BIP, and are modulated by estrogen (E2). However, most of the knowledge we have so far on the neuromodulatory effects of E2 came from studies on animal models and human cell lines, and may not accurately reflect many processes occurring exclusively in the human brain. Thus, here we highlight the advantages of using induced pluripotent stem cell (iPSC) models to revisit studies of mechanisms underlying beneficial effects of E2 in human brain cells. A better understanding of these mechanisms opens the opportunity to identify putative targets of novel therapeutic agents for SCZ and BIP. In this review, we first summarize the literature on the molecular mechanisms involved in SCZ and BIP pathology and the beneficial effects of E2 on neuron–glia interactions. Then, we briefly present the most recent developments in the iPSC field, emphasizing the potential of using patient-derived iPSCs as more relevant models to study the effects of E2 on neuron–glia interactions.
-
Osete, Jordi Requena; Akkouh, Ibrahim; Assis, Denis Reis de; Szabo, Attila; Frei, Evgeniia & Hughes, Timothy
[Show all 10 contributors for this article]
(2021).
Lithium increases mitochondrial respiration in iPSC-derived neural precursor cells from lithium responders.
Molecular Psychiatry.
ISSN 1359-4184.
doi:
10.1038/s41380-021-01164-4.
Full text in Research Archive
Show summary
Lithium (Li), valproate (VPA) and lamotrigine (LTG) are commonly used to treat bipolar disorder (BD). While their clinical efficacy is well established, the mechanisms of action at the molecular level are still incompletely understood. Here we investigated the molecular effects of Li, LTG and VPA treatment in induced pluripotent stem cell (iPSC)-derived neural precursor cells (NPCs) generated from 3 healthy controls (CTRL), 3 affective disorder Li responsive patients (Li-R) and 3 Li non-treated patients (Li-N) after 6 h and 1 week of exposure. Differential expression (DE) analysis after 6 h of treatment revealed a transcriptional signature that was associated with all three drugs and most significantly enriched for ribosome and oxidative phosphorylation (OXPHOS) pathways. In addition to the shared DE genes, we found that Li exposure was associated with 554 genes uniquely regulated in Li-R NPCs and enriched for spliceosome, OXPHOS and thermogenesis pathways. In-depth analysis of the treatment-associated transcripts uncovered a significant decrease in intron retention rate, suggesting that the beneficial influence of these drugs might partly be related to splicing. We examined the mitochondrial respiratory function of the NPCs by exploring the drugs’ effects on oxygen consumption rate (OCR) and glycolytic rate (ECAR). Li improved OCR levels only in Li-R NPCs by enhancing maximal respiration and reserve capacity, while VPA enhanced maximal respiration and reserve capacity in Li-N NPCs. Overall, our findings further support the involvement of mitochondrial functions in the molecular mechanisms of mood stabilizers and suggest novel mechanisms related to the spliceosome, which warrant further investigation.
View all works in NVA
-
-
-
Stavrum, Anne-Kristin; Szabo, Attila; Fernandes, Carla Patricia Duarte; Frei, Evgeniia; Djurovic, Srdjan & Hellard, Stephanie Le
(2020).
Epigenetic Effects of THC and CBD in Neuronal Stem Cells.
Full text in Research Archive
-
Frei, Evgeniia
(2026).
Integrating Genetic, Environmental, and Behavioral Perspectives on Adolescent Mental Health: Population-Based Analyses and Predictive Modelling.
Universitetet i Oslo.
ISSN 9788234808654.
Full text in Research Archive
Show summary
This thesis examined genetic, environmental, and behavioral factors underlying adolescent mental health, with a focus on screen use as a widespread environmental exposure. Data from the Norwegian Mother, Father, and Child Cohort Study (MoBa) were used, combining questionnaire, genetic, and diagnostic information.
The genetic architecture of screen behaviors (TV watching, gaming, social media use) was characterized through genome-wide association studies and post-GWAS analyses. Screen behaviors showed modest heritability, with negative genetic correlations with educational attainment and significant genetic overlaps with psychiatric disorders.
Phenotypic analyses revealed that screen use was linked to elevated odds of psychiatric diagnoses such as depression, anxiety, and neurodevelopmental disorders. Very low social media use was also associated with higher risk, challenging a simple "more is worse" assumption. Genetic sensitivity analyses indicated that associations between high screen use and psychiatric diagnoses are confounded by shared genetic factors.
Data-driven approaches integrated these factors within a predictive framework for adolescent mental health risk. Machine learning models performed best using self-reported symptoms and psychosocial variables, while screen use and genetic factors contributed little. Simplified models using three questionnaire items – e.g. mental distress, sleep problems, and weekly exercise – achieved discrimination comparable to larger models.
Overall, the findings show that screen use and mental health associations are neither uniform nor unidirectional, vary across outcomes and screen behaviors, and are partly shaped by shared genetic factors. This challenges simplistic narratives around screen use and has implications for prevention policy. The results further suggest that brief symptom-based questionnaires can support identification of adolescents at elevated psychiatric risk in practical settings.
View all works in NVA
Published
July 7, 2022 3:14 PM
- Last modified
Nov. 6, 2024 11:58 AM
