Proteonano™ enables deep proteomic analysis of DKD patient serum samples
Figure 1: Sample preprocessing and LC-MS analysis.
1393 protein groups were identified in 67 samples. 1185 of them were mapped to the human plasma protein project protein catalog. Concentrations of these proteins spanned 8 orders of magnitude, with lowest protein concentration of 3.0 pg/mL.
Differentially expressed proteins between between DKD progressors and non-progressors
Figure 2. Differentially expressed proteins in DKD patients with and without disease progression.
45 proteins were differentially expressed between DKD progressors and non-progressors. ROC analysis showed the best per- forming single protein that can differentiate non-progressors and progressors was VWF (AUC was 0.785).
Identification of effective biomarker combinations to predict DKD progression
Figure 3. Assessment of multivariate models.
Multivariate analyses were subsequently carried out. Eight feature selection methods were employed, including least absolute shrinkage and selection operator (LASSO) and random forest (RF). Features of the best-performing model were selected by the RF method, with five proteins in the panel (ROC-AUC=0.97), superior to the discriminative power of UACR (ROC-AUC=0.87).
