type_marker_test_res = top_markers(cds,
group_cells_by="cluster",
reference_cells=1000,
cores=8)
# Require that markers have at least JS specificty score > 0.5 and
# be significant in the logistic test for identifying their cell type:
garnett_markers = type_marker_test_res %>%
filter(marker_test_q_value < 0.01 & specificity >= 0.5) %>%
group_by(cell_group) %>%
top_n(5, marker_score)
# Exclude genes that are good markers for more than one cell type:
garnett_markers = garnett_markers %>% group_by(gene_short_name) %>%
filter(n() == 1)
generate_garnett_marker_file(garnett_markers, file="~\marker_file.txt")
class <- train_cell_classifier(cds = cds,
marker_file =
"~\marker_file.txt",
db=org.Mm.eg.db::org.Mm.eg.db,
num_unknown = 50,
marker_file_gene_id_type = "SYMBOL",
cores=8)