Project test-set samples onto training NMF factors
project_test_factors.RdConvenience wrapper that combines compute.test.set() and
project_factors() into a single call. Given a trained
CellTFusion() result and a test deconvolution matrix, it
(1) reconstructs cell group composite scores for the test samples using
the training CCA loadings, and (2) projects those scores onto the trained
NMF latent space.
Arguments
- train_processed
A list returned by
CellTFusion(), containing at minimumProcessed_deconvolution,Cell_groups, andLatent_spaces.- test_deconv
A numeric matrix or data frame of deconvolution features for the test samples (samples x cell types). Column names must match those used during training. If the model was trained with batch correction, the test samples are centred on their own means (as each training cohort was), so they should come from a single cohort.