In their research, the Nolan Lab uses metabolomics of World Trade Center lung injury (WTC-LI) and machine learning. From a rich dataset of hundreds of variables, only those that maximally highlight class separation are chosen. What we are left with is an optimal subset of predictive metabolites that elucidate key phenotypic variations related to case status.

Linear combinations of key metabolites comprise principal components which are used to demonstrate class separation produced by the refined metabolite profile. Cases of WTC-LI are clearly distinguishable from controls based on these scores. These key metabolites are then combined with previously identified, clinically significant serum and clinical biomarkers in an integrated multiplatform analysis to create a predictive biomarker profile with clinical utility.