kBET: kBET - k-nearest neighbour batch effect test

In theislab/kBET: k-nearest neighbour batch effect test

kBET - k-nearest neighbour batch effect test

Description

kBET runs a chi square test to evaluate the probability of a batch effect.

Usage

kBET(df, batch, k0 = NULL, knn = NULL, testSize = NULL,
  do.pca = TRUE, dim.pca = 50, heuristic = TRUE, n_repeat = 100,
  alpha = 0.05, addTest = FALSE, verbose = FALSE, plot = TRUE,
  adapt = TRUE)

Arguments

df	dataset (rows: cells, columns: features)
batch	batch id for each cell or a data frame with both condition and replicates
k0	number of nearest neighbours to test on (neighbourhood size)
knn	an n x k matrix of nearest neighbours for each cell (optional)
testSize	number of data points to test, (10 percent sample size default, but at least 25)
do.pca	perform a pca prior to knn search? (defaults to TRUE)
dim.pca	if do.pca=TRUE, choose the number of dimensions to consider (defaults to 50)
heuristic	compute an optimal neighbourhood size k (defaults to TRUE)
n_repeat	to create a statistics on batch estimates, evaluate 'n_repeat' subsets
alpha	significance level
addTest	perform an LRT-approximation to the multinomial test AND a multinomial exact test (if appropriate)
verbose	displays stages of current computation (defaults to FALSE)
plot	if stats > 10, then a boxplot of the resulting rejection rates is created
adapt	In some cases, a number of cells do not contribute to any neighbourhood and this may cause an imbalance in observed and expected batch label frequencies. Frequencies will be adapted if adapt=TRUE (default).

Value

list object

1. summary - a rejection rate for the data, an expected rejection rate for random labeling and the significance for the observed result

2. results - detailed list for each tested cells; p-values for expected and observed label distribution

3. average.pval - significance level over the averaged batch label distribution in all neighbourhoods

4. stats - extended test summary for every sample

5. params - list of input parameters and adapted parameters, respectively

6. outsider - only shown if adapt=TRUE. List of samples without mutual nearest neighbour:

   • index - index of each outsider sample)

   • categories - tabularised labels of outsiders

   • p.val - Significance level of outsider batch label distribution vs expected frequencies. If the significance level is lower than alpha, expected frequencies will be adapted

If the optimal neighbourhood size (k0) is smaller than 10, NA is returned.

Examples

    batch <- rep(seq_len(10),each=20)
    data <- matrix(rpois(n = 50000, lambda = 10)*rbinom(50000,1,prob=0.5), nrow=200)

    batch.estimate <- kBET(data,batch)

theislab/kBET documentation built on Jan. 27, 2024, 9:58 p.m.