Development of an integrative cross-omics approach for conceptual adverse outcome pathway network construction

The abilities of recent high throughput techniques to measure biological responses is rapidly growing, therefore methods to analyse and organise these vast amounts of data into meaningful results are needed. Adverse outcome pathways (AOPs) and AOP networks (AOPNs) are an increasingly recognised framework for translating mechanistic information into useable knowledge to support policy decisions. However, many traditional statistical approaches may be ineffective at capturing nuances of high throughput data, particularly from multiple disparate layers of biological organisation. We present a comprehensive method that combines univariate differential expression (UD) analysis and multivariate integrative modeling (MIM) approaches, using transcriptomic and metabolomic data from adipocytes exposed to a classic obesogen, to develop a conceptual AOPN (cAOPN) for metabolic syndrome (MetS). Simpson-Golabi-Behmel syndrome (SGBS) preadipocyte cells were differentiated in tributyltin (TBT) and analysed using whole genome transcriptome and untargeted metabolomics analysis. UD and MIM results were used to identify perturbed features (PFs) for over-representation analysis for pathways and diseases and followed by integrated network and cluster analyses based on Jaccard similarity to reorganise resultant complex biological phenomena into exploratory depictions of cause-and-effect relationships. The resulting cAOPN for MetS was assembled and corroborated with the literature and mechanistic pathway databases that supported the identified disruptions in lipid regulation, iron transport, growth processes, key signalling processes, adipocyte differentiation, and hormonal homeostasis. Overall, by leveraging the strengths of multiple statistical methods in combination with heterogeneous data from multiple layers of biological organisation, this method facilitated the integration and interpretation of complex data into an exploratory mechanistic schema for AOP and AOPN hypothesis generation and prioritisation.