Multi-Scale Biological Data Integration

Abstract

Biological systems operate across scales from molecular interactions to cellular phenotypes, tissue organisation, and organismal physiology, and integrating data across these scales is essential for understanding emergent properties that single-scale analyses miss, yet adoption of multi-scale integration methods varies across methodological approaches. We evaluated 214 multi-scale biological data integration programmes across centres in Italy, France, and Sweden between 2018 and 2022, spanning five integration categories: multi-omics factor analysis, cross-scale network integration, spatial multi-modal alignment, hierarchical Bayesian multi-scale models, and deep multi-view representation learning. A Multi-Scale Integration Index (MSII) was constructed from five sub-scores -- cross-scale prediction accuracy, biological coherence of integrated features, scalability to high-dimensional multi-modal data, interpretability of latent factors, and novel biological insight yield -- with weights from regression against sustained adoption. MSII correlated with adoption at r = +0.84 and discriminated adopted from non-adopted methods with an AUC of 0.882. Multi-omics factor analysis scored highest (mean MSII 0.824), while hierarchical Bayesian models trailed at 0.598. Only 35.5 percent exceeded the 0.75 threshold. Cross-scale accuracy carried the largest weight (beta = +0.278), followed by biological coherence (beta = +0.230)