Models can be loaded into scGate using the syntax:
models.db <- get_scGateDB()
model.Tcell <- models.db$human$generic$Tcell
gated <- scGate(obj, model=model.Tcell)To use scGate as a multi-classifier, load one of the collections from the DB:
models.db <- get_scGateDB()
models.TME <- models.db$human$TME_HiRes
gated <- scGate(obj, model=models.TME)
table(gated$scGate_multi)Note: we will tag versions of this DB using the format 'vN.M' (e.g. v1.4). This format is understood by scGate to download specific versions of the DB
For common use cases, there are ready-made model collections available. For example, there is a model collection to classify immune cells in the tumor micro-environment at high resolution:
scGate_models_DB <- get_scGateDB()
scGate_models <- scGate_models_DB$human$TME_HiResModel collections available:
-
HiTME
Default model used in the HiTME package. Includes immune cells at high resolution and stromal cells (e.g. fibroblasts). -
PBMC
Includes immune cells at high resolution, optimized for possible background contamination from lysed blood cells. -
TME_broad
Classification of stromal cells and immune cells at a broad level.
Changelog for model collections:
- PBMC
- For all cells, monocytes negative levels replaced with MoMacDC negative levels due to better specificity
- Monocytes: Removed NK negative level because some Monocytes also express NKG7
- Monocytes: kept only "myeloid" markers, as SPI1 is most specific for Monocytes
- human/generic/MAIT
Two levels: (T cells or MAIT-like cells) minus epithelial, stromal, endothelial, B, myeloid and plasma cells, then MAIT (SLC4A10, TRAV1-2, KLRB1, IL18RAP, CCR6) minus gamma-delta T and NK cells. Designed to work from whole tissue as well as blood. Run it withparam_decay = 0: the default shrinks the neighbourhood and gene set at each level, which blurs rare MAIT cells into other memory T cells.MAIT.level1has the same genes asMAITbut a separate name, so the level-2 score is not carried over from level 1.