Welcome to the MultiMuTHER study
MultiMuTHER is a longitudinal multi-omic study designed to investigate how molecular traits - such as gene expression and serum metabolite levels - change over time within individuals. The study includes 335 female participants from the TwinsUK registry, each of whom has provided samples at three or more clinical visits.
The cohort includes both monozygotic and dizygotic twins. Multi-omic profiling currently consists of whole-blood bulk RNA sequencing and metabolomic profiling on the Metabolon platform, complemented by extensive genotypic, clinical, biochemical, and anthropometric data. Together, these datasets provide a rich resource for understanding molecular changes over time.
Summary statistics for all genes from longitudinal gene expression association analysis of gene expression levels against time from baseline visit in the MultiMuTHER study. Betas and standard errors for each gene are for the time from baseline visit fixed effect in a longitudinal random slope and random intercept mixed effects model (see methods). P-values for each gene correspond to the time from visit fixed effect (P-value Fixed), likelihood ratio tests comparing mixed effects models including both a fixed effect and random slope for time from baseline to a random intercept only model (P-value RandomSlope), and a model including only a random slope for time from baseline (no fixed effect) to a null random intercept model where time from baseline was removed (P-value RandomSlopeOnly). 5% false discovery rates for the corresponding tests were employed to classify genes as showing population level trends in change over time (BH-adjusted P Fixed) or individual-specific change over time (both BH-adjusted P RandomSlope and BH-adjusted P RandomSlopeOnly).
Summary statistics for longitudinal time from baseline x age at baseline visit interaction gene expression association analysis in the MultiMuTHER study. The effect sizes, standard errors, p-values and BH-adjusted p-values correspond to the time from baseline x age at baseline visit interaction term. A 5% FDR was employed to classify genes as showing differences in longitudinal trajectories by age.
Summary statistics for all genes from longitudinal gene expression association analysis of gene expression levels against time from baseline visit using cell-type specific gene expression levels in the MultiMuTHER study. Cell-type specific expression levels were inferred from bulk whole blood RNA-seq data using CIBERSORTx. Summary statistics are provided for each of the inferred cell types. Betas and standard errors for each gene are for the time from baseline visit fixed effect in a longitudinal random intercept mixed effects model (see methods). P-values for each gene correspond to the time from visit fixed effect (P-value). 5% false discovery rates for the corresponding tests were employed to identify longitudinally-variable genes in each cell type. Genes whose cell-type specific inferred expression did not pass the expression variance threshold are indicated with NAs for the relevant cell type.
Summary statistics for longitudinal genotype by time cis-eQTL interaction analyses. Longitudinal linear mixed effects models included an interaction term for genotype dosage at each assessed variant and time from baseline visit. Variant-gene pairs with genotype x time cis-eQTL interaction significant at a 5% Benjamini-Hochberg FDR are shown. P-values correspond to the genotype x time interaction term.
Summary statistics for gene expression association analyses of gene expression levels against time of visit in the MultiMuTHER study. Linear mixed effects models were used to assess association with time of clinical visit (5% FDR).
Summary statistics for gene expression association analyses of gene expression levels against seasonality in the MultiMuTHER study. Cosinor linear mixed effects models were used to assess association with seasonality (5% FDR).
Summary statistics for gene expression association analyses of gene expression levels against serum PFOA levels in the MultiMuTHER study. Association results are provided from linear mixed effects models including cross-sectional data from either the first or last clinical visits, at 5% FDR.
Summary statistics for gene expression association analyses of gene expression levels against serum PFOS levels in the MultiMuTHER study. Association results are provided from linear mixed effects models including cross-sectional data from either the first or last clinical visits, at 5% FDR.
Summary statistics for all metabolites from longitudinal metabolome-wide association analysis of serum metabolite levels against time from baseline visit in the MultiMuTHER study. Betas and standard errors for each metabolite are for the time from baseline visit fixed effect in a longitudinal random slope and random intercept mixed effects model (see methods). P-values for each metabolite correspond to the time from visit fixed effect (P-value Fixed), likelihood ratio tests comparing mixed effects models including both a fixed effect and random slope for time from baseline to a random intercept only model (P-value RandomSlope), and a model including only a random slope for time from baseline (no fixed effect) to a null random intercept model where time from baseline was removed (P-value RandomSlopeOnly). 5% false discovery rates for the corresponding tests were employed to classify metabolites as showing population level trends in change over time (BH-adjusted P Fixed) or individual-specific change over time (both BH-adjusted P RandomSlope and BH adjusted P RandomSlopeOnly).
Summary statistics for longitudinal time from baseline x age at baseline visit interaction metabolome-wide association analysis of serum metabolite levels in the MultiMuTHER study. The effect sizes, standard errors, p-values and BH-adjusted p-values correspond to the time from baseline x age at baseline visit interaction term. A 5% FDR was employed to classify metabolites as showing differences in longitudinal trajectories by age.
Summary statistics for genome-wide association studies of longitudinal rates of change of metabolite levels. Linear mixed effects models for individual longitudinal slopes for each metabolite were fitted genomewide for 909/915 metabolites, with variant-metabolite pairs significant at P < 1.8 x 10-10 shown.
Summary statistics for metabolome-wide association analyses of serum metabolite levels against time of visit in the MultiMuTHER study. Linear mixed effects models were used to assess association with time of clinical visit (5% FDR).
Summary statistics for metabolome-wide association analyses of serum metabolite levels against seasonality in the MultiMuTHER study. Cosinor linear mixed effects models were used to assess association with seasonality (5% FDR).
Summary statistics for metabolome-wide association analyses of serum metabolite levels against PFOA levels in the MultiMuTHER study. Association results are provided from linear mixed effects models including cross-sectional data from either the first or last clinical visits, at 5% FDR.
Summary statistics for metabolome-wide association analyses of serum metabolite levels against PFOS levels in the MultiMuTHER study. Association results are provided from linear mixed effects models including cross-sectional data from either the first or last clinical visits, at 5% FDR.
Summary statistics for association analyses between whole blood gene expression and serum metabolite levels in the MultiMuTHER study. Gene-metabolite pairs at BH-adjusted p-value ≤ 0.05 are shown. N denotes the total number of observations included in the model, pooled across all participants and all available longitudinal timepoints.