Phospholipids are important to cellular function and are a vital structural component of plasma and organelle membranes. These membranes isolate the cell from its environment, allow regulation of the internal concentrations of ions and small molecules, and host diverse types of membrane proteins. It remains extremely challenging to identify specific membrane protein–lipid interactions and their relative strengths. Native mass spectrometry, an intrinsically gas-phase method, has recently been demonstrated as a promising tool for identifying endogenous protein–lipid interactions. However, to what extent the identified interactions reflect solution- versus gas-phase binding strengths is not known. Here, the “Extended” Kinetic Method and ab initio computations at three different levels of theory are used to experimentally and theoretically determine intrinsic gas-phase basicities (GB, ΔG for deprotonation of the protonated base) and proton affinities (PA, ΔH for deprotonation of the protonated base) of six lipids representing common phospholipid types. Gas-phase acidities (ΔG and ΔH for deprotonation) of neutral phospholipids are also evaluated computationally and ranked experimentally. Intriguingly, it is found that two of these phospholipids, sphingomyelin and phosphatidylcholine, have the highest GB of any small, monomeric biomolecules measured to date and are more basic than arginine. Phosphatidylethanolamine and phosphatidylserine are found to be similar in GB to basic amino acids lysine and histidine, and phosphatidic acid and phosphatidylglycerol are the least basic of the six lipid types studied, though still more basic than alanine. Kinetic Method experiments and theory show that the gas-phase acidities of these phospholipids are high but less extreme than their GB values, with phosphatidylserine and phosphatidylglycerol being the most acidic. These results indicate that sphingomyelin and phosphatidylcholine lipids can act as charge-reducing agents when dissociated from native membrane protein–lipid complexes in the gas phase and provide a straightforward model to explain the results of several recent native mass spectrometry studies of protein–lipid complexes.

The use of machine learning (ML) with metabolomics provides opportunities for the early diagnosis of disease. However, the accuracy of ML and extent of information obtained from metabolomics can be limited owing to challenges associated with interpreting disease prediction models and analyzing many chemical features with abundances that are correlated and “noisy”. Here, we report an interpretable neural network (NN) framework to accurately predict disease and identify significant biomarkers using whole metabolomics data sets without a priori feature selection. The performance of the NN approach for predicting Parkinson’s disease (PD) from blood plasma metabolomics data is significantly higher than other ML methods with a mean area under the curve of >0.995. PD-specific markers that predate clinical PD diagnosis and contribute significantly to early disease prediction were identified including an exogenous polyfluoroalkyl substance. It is anticipated that this accurate and interpretable NN-based approach can improve diagnostic performance for many diseases using metabolomics and other untargeted ‘omics methods.