ElasticNet and PLS-R
To identify the key contributing antibody signatures from the data set,
elasticNet regression was utilised as previously
described19, 38 The Elastic Net hyperparameter was set
to have equal weights between L1norm and L2norm i.e. α=0.5. Model
performance was evaluated iteratively (1000 iterations, 4 fold cross
validation). Partial least squares regression (PLS-R) was performed on
ElasticNet-selected antibody features to visualize the relationship
between antibody signatures with continuous variables. i.e. Ct (cycle
threshold) PCR values (determined via Roche LightCycler 480 Instrument
II).