Celia M.T. Greenwood

The paper develops FunBurd to analyze gene dosage effects on 43 traits in UK-Biobank, finding pleiotropy and non-monotonic responses linked to genetic constraint and brain-specific mechanisms.

MOLA identifies multiomic loops in datasets, showing better capture of epigenetic alterations in Influenza A virus data and associations with progression in breast cancer.

The paper investigates how sample size and cohort composition affect the replicability of brain-behaviour associations using Canonical Correlation Analysis on UK Biobank data.

The paper develops a statistical method to detect locus-specific escape from X-chromosome inactivation in DNA methylation data, applied to rheumatoid arthritis and venous thromboembolism datasets.

This paper evaluates classifiers for detecting causal language in scientific abstracts and applies the best model to analyze trends in microbiome research from 2015 to 2025.

The paper investigates five BRIP1/FANCJ variants for their clinical relevance in breast and ovarian cancer, finding three likely damaging variants associated with cellular sensitivity to chemotherapy drugs.

The paper analyzes germline DNA repair gene variants in BRCA1 and BRCA2 families to assess the prevalence of damaging variants in DNA repair genes beyond established breast cancer predisposing genes.

The paper develops a method to analyze genomic deletions and duplications across the cerebral cortex and finds opposing effects on cognitive ability depending on gene expression patterns in different cortical regions.

The paper introduces a Bayesian functional regression model, DMCFB, to identify differentially methylated cytosines in cancer epigenetics, showing improved performance over existing methods.

This study evaluates AlphaMissense's impact on gene-based testing for disease causation using UK Biobank exome data, finding no significant improvement in identifying disease-causing genes.

The paper introduces a method called sail to detect non-linear interactions between a key exposure variable and high-dimensional data, showing it outperforms existing methods in prediction and support recovery.

The paper develops a latent factor model to capture additional genetic risk from family history, improving polygenic risk prediction for height and 11 complex diseases.

The paper introduces the Effector Index, an algorithm to predict causal genes at GWAS loci using genomic features, achieving 85% AUC in type 2 diabetes validation.

This paper identifies FANCI c.1813C>T as a candidate familial ovarian cancer risk allele in a French Canadian population with founder effects.

Neuropsychiatric CNVs reorganize brain connectivity along shared patterns linked to cognitive traits and disorder severity.

This study shows that early gestational MIA exposure in mice leads to lasting brain and behavioral changes, while late exposure has less impact, with molecular changes in the dorsal hippocampus.

This study identifies OAS1 as a protective factor against COVID-19 severity and susceptibility using Mendelian randomization and case-control analyses.

This study evaluates how different preprocessing pipelines affect neuroimaging analyses, showing that software, parcellation, and quality control significantly impact results.

This paper assesses the statistical power of case-control genomic studies for identifying host genetic factors associated with COVID-19 susceptibility and severity using simulations and current epidemiological estimates.

The paper estimates the effect sizes of deletions and duplications on autism risk and IQ using statistical models applied to CNV data from multiple cohorts.

This paper introduces ggmix, a penalized linear mixed model for SNP selection and population structure adjustment in high-dimensional genetic prediction models, showing improved accuracy and parsimony over existing methods.

The paper quantifies the effect of copy number variants on cognitive ability using genomic data from 24,092 individuals, finding that 10,000 genes modulate intelligence.

MDiNE estimates differential co-occurrence networks in microbiome studies to analyze how microbial networks change with a binary covariate, outperforming existing methods in simulations and applied to Crohn's disease.

The paper identifies 9 novel genetic loci, including low-frequency variants in TP53, associated with head circumference and intracranial volume in up to 46,000 individuals.

This paper introduces Constrained Instrumental Variable (CIV) methods to improve Mendelian randomization by handling pleiotropy, showing better performance in simulations and application to ADNI data.

This paper uses machine learning to predict osteoporotic fracture risk from genotypes, finding that genomic predictions outperform clinical risk factors in identifying fracture risk.

PRS-on-Spark generates polygenic risk scores efficiently, supporting diverse data inputs and improving variance explained for major depression.

The paper introduces ECLUST, a framework that improves predictive modeling in high-dimensional data by incorporating exposure-dependent clustering, showing better performance than traditional methods.