Abstract
Characterising the metabolic response to an intervention can be challenging due to the large inter-subject variation. Because traditional chemometric approaches such as OPLS -DA do not take into account the paired data structure in these studies, we have adopted multivariate paired data analysis (MVPDA) to enhance the recovery of metabolic biomarkers from sheep subjected to surgically-induced osteoarthritis (OA). Sheep underwent one of three types of surgical procedure (sham (control), meniscal destabilisation, MD or anterior cruciate ligament transaction, ACLT), and for every animal a serum sample was collected prior to being operated on and at sacrifice. 1D 1H NMR spectra were acquired from each sample at 800 MHz. Results from traditional chemometric techniques (PCA, OPLS-DA) were compared and contrasted with MVPDA, which displayed enhanced classification and interpretability. MVPDA showed all types of surgical procedure were associated with elevated lactate and deceased TMAO/betaine. Serum from sheep that underwent ACLT was additionally characterised by elevated branched-chain amino acids (BCAAs) and decreased histidine, consistent with previous findings. There was no observable change in BCAAs for the MD cohort indicating that different OA subtypes were associated with unique metabolic fingerprints. This study reinforces the utility of MVPDA for datasets with a paired structure, and offers new insights into the metabolic consequences of OA.