Reference API

AbstractGridCounts

A stack of genes each consisting of a 2D count array of the same size.

GridCounts{G,T<:Number} <: AbstractGridCounts

A stack of genes::G each consisting of a 2D count::T array of the same size.

GridCounts(df::DataFrame, binsize::Real)

Construct from a DataFrame with columns 'x' (Real), 'y' (Real), 'count', and 'geneID' (Pool) binning the data by binsize.

GridCounts(df::DataFrame)

Construct from a DataFrame with columns 'x' (Integer), 'y' (Integer), 'count', and 'geneID' (Pool).

GridCounts(counts::Dict)

All values of the dict must be arrays of the same size.

crop!(counts, slice)

Crop each gene layer in counts by indexing with slice.

gridsize(counts)

The size of the grid of each layer.

mask!(counts, mask::AbstractMatrix{Bool})

Remove all counts in each gene layer for which mask is false.

totalrna(counts)

Caclulate the totalrna as sum of all genes for each pixel.

readGEMfile(file)

Read file in GEM format as DataFrames.DataFrame.

readstereoseq(file)

Read StereoSeq file as GridCounts.

readstereoseqbinned(file, binsize::Integer) -> (counts, genes, bins)

Read StereoSeq file and aggregate into bins.

assigncelltype(counts, signatures, kernel; celltypes=nothing, log=false) -> (celltypes, cosine)

Assign a celltype to each pixel.

The cosine similarity is calculated using signatures to assign the celltype with the highest similarity to each pixel.

The eltype(kernel) will be used for calculations and signatures will be cast to it.

Arguments

• signatures::AbstractDataFrame: celltype signatures (celltypes x genes).
• celltypes::Vector{AbstractString}=nothing: optional celltype names.
• log::Bool: whether to log-transform the KDE. Useful if signatures are calculated from log-transformed gene expression.

gaussiankernel(bw::Real, r::Real)

Generate a gaussian kernel with bandwidth bw and radius r * bw

kde(counts, kernel)

Calculate kernel density estimate.

Arguments

• kernel: usually a centered OffsetArrays.OffsetArray.

findlocalmaxima(img, mindist::Integer; threshold::Real=0)

Find local maxima of the img.

The input should be a kde of the totalrna or comparable.

getlocalmaxima(counts, localmax, kernel; genes=nothing) -> (expression, genes, coordinates)

Load KDE with kernel for coordinates at localmax.

Arguments

• localmax: vector of locations (indices).
• genes=nothing: vector of genes for which to calculate KDE. If nothing all genes are used.