useR! 2024

Membrane balance relies on specific metabolic lipid interactions. We investigated how fatty acyl side chains from different lipid classes influence each other using artificial neural networks (ANNs). We analyzed profiles of different phospholipids from 15 mouse tissues and found tissue-specific patterns. With this data we trained 362 ANN model architectures in R and were able to predict mitochondrial cardiolipin remodelling based on fatty acyl pools in phospholipids. Our analysis revealed key players in mammalian cardiolipin remodelling: high oleic acid increased lipid diversity, while linoleic acid favoured uniformity. Additionally, to reliably discriminate between plasmanyl/plasmenyl lipids in mouse tissues we conducted lipidomics experiments with ion mobility spectrometry (IMS). All data integration and analysis steps were performed in R. Statistical analysis in R confirmed the validity of this IMS approach for lipid subclass separation, which is especially powerful when combined with accurate retention time characteristics.