ProjectUMAP: Project query into UMAP coordinates of a reference

Description

This function will take a query dataset and project it into the coordinates of a provided reference UMAP. This is essentially a wrapper around two steps:

FindNeighbors - Find the nearest reference cell neighbors and their distances for each query cell.
RunUMAP - Perform umap projection by providing the neighbor set calculated above and the umap model previously computed in the reference.

Usage

ProjectUMAP(query, ...)
# S3 method for default
ProjectUMAP(
  query,
  query.dims = NULL,
  reference,
  reference.dims = NULL,
  k.param = 30,
  nn.method = "annoy",
  n.trees = 50,
  annoy.metric = "cosine",
  l2.norm = FALSE,
  cache.index = TRUE,
  index = NULL,
  neighbor.name = "query_ref.nn",
  reduction.model,
  ...
)
# S3 method for DimReduc
ProjectUMAP(
  query,
  query.dims = NULL,
  reference,
  reference.dims = NULL,
  k.param = 30,
  nn.method = "annoy",
  n.trees = 50,
  annoy.metric = "cosine",
  l2.norm = FALSE,
  cache.index = TRUE,
  index = NULL,
  neighbor.name = "query_ref.nn",
  reduction.model,
  ...
)
# S3 method for Seurat
ProjectUMAP(
  query,
  query.reduction,
  query.dims = NULL,
  reference,
  reference.reduction,
  reference.dims = NULL,
  k.param = 30,
  nn.method = "annoy",
  n.trees = 50,
  annoy.metric = "cosine",
  l2.norm = FALSE,
  cache.index = TRUE,
  index = NULL,
  neighbor.name = "query_ref.nn",
  reduction.model,
  reduction.name = "ref.umap",
  reduction.key = "refUMAP_",
  ...
)

Arguments

query: Query dataset
...: Additional parameters to RunUMAP
query.dims: Dimensions (columns) to use from query
reference: Reference dataset
reference.dims: Dimensions (columns) to use from reference
k.param: Defines k for the k-nearest neighbor algorithm
nn.method: Method for nearest neighbor finding. Options include: rann, annoy
n.trees: More trees gives higher precision when using annoy approximate nearest neighbor search
annoy.metric: Distance metric for annoy. Options include: euclidean, cosine, manhattan, and hamming
l2.norm: Take L2Norm of the data
cache.index: Include cached index in returned Neighbor object (only relevant if return.neighbor = TRUE)
index: Precomputed index. Useful if querying new data against existing index to avoid recomputing.
neighbor.name: Name to store neighbor information in the query
reduction.model: DimReduc object that contains the umap model
query.reduction: Name of reduction to use from the query for neighbor finding
reference.reduction: Name of reduction to use from the reference for neighbor finding
reduction.name: Name of projected UMAP to store in the query
reduction.key: Value for the projected UMAP key