Fig. 2: Integrated analysis of endometrium using multi-omics, including transcriptomics, proteomics, and targeted genomic sequencing.
a Left: The relative contribution of the transcriptomic and proteomic datasets to MOFA-inferred factors, expressed as the percentage of explained variance, with intensity represented in blue. Right: Correlation of factor variance with clinical and genetic parameters, quantified by -log10 (adjusted p-value) and visualised in red. Parameters include experimental batches (n = 3), fibroid presence (UF vs non-UF), hormone treatment (past or current), heavy menstrual bleeding (HMB) status, and fibroid-associated mutations: canonical MED12 UF mutations, COL4A6 rs6622312, AHR rs2066853 and FH rs6673988. b Scatter plots illustrating the differentiation of samples based on key clinical and genetic parameters, including HMB (Yes, n = 10; No, n = 21), hormone past (prior hormone treatment: Yes, n = 11; No, n = 16; Unknown, n = 4), hormone current (treatment at time of surgery: Yes, n = 8; No, n = 20; Unknown, n = 3), fibroid (UF, n = 23; non-UF, n = 8), MED12 UF mutations (wt, n = 14; mut, n = 9; non-fibroid, n = 8), COL4A6 rs6622312 (wt, n = 12; mut, n = 11; non-fibroid, n = 8), and AHR rs2066853 (wt, n = 16; mut, n = 7; non- fibroid, n = 8). MOFA factor values represent the relative positioning of samples, with larger absolute values indicating stronger associations. c Boxplots showing the distribution of sample groups across MOFA factors 1, 2, and 7, revealing variance within these factors. The centre line represents the median; boxes represent the interquartile range (IQR), and whiskers extend to 1.5 times of IQR. d Gene ontology (GO) enrichment analysis highlighting pathways of features contributing to Factor 1 in both omics (FDR < 0.1). e STRING network diagrams elucidating the interactions among features associated with Factor 1 in both modalities (absolute loading weight higher than 0.3). The loading weight of each feature was identified by MOFA using Bayesian framework and sparsity-induced priors, different from classical regression using p-values for significance. Only relevant features have non-zero loading weight. f GO enrichment analysis of features contributing to Factor 7 in both omics (FDR < 0.1). g STRING network diagrams of features associated with Factor 7 in both modalities (absolute loading weight higher than 0.3).
