Biomedical Data Analytics

Integration across genomics, transcriptomics, proteomics and metabolomics data layers using latent factor methods (MOFA+), multi-block partial least squares (mixOmics), and network-based integration. Analysis identifies cross-layer regulatory relationships, shared variance, and multi-omic biomarker signatures. Outputs include factor loadings, variance explained per layer, and network-visualised cross-modal associations.

Hypothesis testing and regression modelling for clinical trial, observational study, and EHR-derived datasets. Methods: parametric and non-parametric tests, linear and logistic regression, Poisson and negative binomial models, linear mixed-effects models (lme4), and propensity score matching/weighting for causal inference. Multiple testing correction (Bonferroni, BH-FDR) applied throughout. Outputs include statistical reports with effect sizes and confidence intervals formatted for manuscript or regulatory submission.

Kaplan-Meier estimation with log-rank tests, Cox proportional hazards models with assumption checking (Schoenfeld residuals, log-log plots), time-varying covariate extensions, and competing risk analysis using the Fine-Gray model. Parametric survival models (Weibull, log-normal, log-logistic) fitted where proportional hazards assumptions are violated. Commonly used for overall survival, progression-free survival, and recurrence analysis in oncology and cardiovascular studies.

Discovery and validation of diagnostic or prognostic biomarkers. Statistical approaches: univariate screening with FDR correction, LASSO/elastic-net regularised regression for feature selection, and multivariate panel construction. Diagnostic performance assessed by ROC analysis (AUC, sensitivity, specificity, likelihood ratios), with bootstrapped 95% CIs. Validation in independent held-out cohorts or by k-fold cross-validation with optimism correction.

Production of publication-quality figures for biological and clinical data, formatted to journal specifications (300 dpi, vector or TIFF). Figure types: heatmaps (ComplexHeatmap), volcano plots, MA plots, forest plots, Kaplan-Meier curves, waterfall plots, PCA/UMAP scatter plots, network diagrams, and multi-panel composite figures. Figures are provided as editable vector files (PDF/SVG) and high-resolution rasters, with R or Python scripts for reproducibility.