Background
Lipids are critical biomolecules and high-resolution lipidomics is useful for the evaluation of their pathways and their response to adverse events and treatment. Chromatography combined with mass spectrometry allows for the measurement of relative proportions of specific lipids. While this is a well-established technique, the synthesis of multiple untargeted analysis data streams remains an obstacle for high throughput systems biology.
Approach
The arachidonic acid pathway was used as a model. Lyophilized (powdered) brain tissue from Sprague Dawley rats was fortified with compounds specific to the arachidonic acid metabolic pathway at varying concentrations (18 total). The samples were extracted and analyzed by Liquid Chromatography (LC) - Mass Spectrometry (MS) along with non-fortified negative controls. Multiple instruments and methodologies were applied to obtain traditional, ground truth data for comparison to a high throughput, untargeted analysis approach. We tested our in-house machine learning based platform for mass spectrometry data for its ability to autonomously process high throughput, untargeted lipidomics.
Accomplishments
The automated approach with SwRI’s Lighthouse software suite generated estimates of abundance for target compounds and log fold changes over native with high test-retest reliability (ICC=0.89 and ICC=0.9 respectively). Lighthouse software was able to speciate all target compounds using an updated arachidonic acid pathway library. Compound identification is confounded by the presence of many isomers, but resolvable using fragmentation.
Untargeted analysis used covariation in abundance across spiked samples to speciate target compounds with 100% accuracy. Furthermore, this covariation across samples identified incidentally co-spiked compounds present in the Laboratory Control Standard (LCS) (F1= 72%, precision= 92.5%, specificity =98.8%, sensitivity=59%). This work expanded the capability of our Lighthouse software/Highlight platform to deduce complex lipidomic pathways based on differences in concentrations between cohorts.