predict_cell: predict_cell

In hemberg-lab/sctransfermap: sctransfermap for predicting cell type in single-cell data (RNA-seq and ATAC-seq)

Description

make a confusion matrix of query data by ZSL with consensus manner

Usage

predict_cell(iscell, query, simthr = Inf, gn_overlap_thr = 0.5,
  R_output = TRUE)

Arguments

iscell	attribute set
query	signature object of query (the result of create_signature)
simthr	default simthreshold for prediction (default: Inf -> considers only simthr in attribute set)
gn_overlap_thr	minimum gene overlap ratio of attribute in attribute set between query data
R_output	TRUE if the result of the function should be R object. FALSE if the result of the function should be julia object

Value

a confusion matrix of prediction result

• row: query cell type or cluster name

• column: cell types in side-information (from query)

• first column: unassign count

Examples

d <- make_input(matrix, train = TRUE)
w2v <- make_w2v("Pubmed_index")
annot <- make_cellontology("cell ontology file")
stopwords <- make_stopwords(stopwords)
signature <- create_signature("user", "Heart", d$norm_matirx, d$rowname, d$cellname, annot, "cell_ontology", w2v_index = TRUE)
confusion <- predict_cell(att_set, signature)
- att_set: pre-existing attribute set

hemberg-lab/sctransfermap documentation built on Nov. 11, 2019, 6:23 a.m.