compute_lisi: Compute Local Inverse Simpson's Index (LISI)

In immunogenomics/LISI: Local Inverse Simpson Index (LISI) for scRNAseq data

Description

Use this function to compute LISI scores of one or more labels.

Usage

compute_lisi(X, meta_data, label_colnames, perplexity = 30, nn_eps = 0)

Arguments

X	A matrix with cells (rows) and features (columns).
meta_data	A data frame with one row per cell.
label_colnames	Which variables to compute LISI for.
perplexity	The effective number of each cell's neighbors.
nn_eps	Error bound for nearest neighbor search with RANN:nn2(). Default of 0.0 implies exact nearest neighbor search.

Value

A data frame of LISI values. Each row is a cell and each column is a different label variable.

Examples

## Example with 400 cells. 
library(lisi)
library(dplyr)
library(tidyr)
library(ggplot2)
library(magrittr)

head(lisi::meta_data)

## Let's color cells by labels. For label 1, there are mixed and non-mixed
## groups. For label 2, all cells are well mixed. 
lisi::X %>% 
  cbind(lisi::meta_data) %>% 
  sample_frac(1L, FALSE) %>% 
  gather(key, val, label1, label2) %>% 
  ggplot(aes(X1, X2, color = val)) +
    geom_point(shape = 21) + 
    facet_wrap(~key)
  
## Now to compute and plot the LISI values for each label. 
lisi_res <- compute_lisi(lisi::X, lisi::meta_data, c('label1', 'label2'))
head(lisi_res)

lisi::X %>% 
  cbind(lisi_res) %>% 
  sample_frac(1L, FALSE) %>% 
  gather(key, lisi_value, label1, label2) %>% 
  ggplot(aes(X1, X2, color = lisi_value)) +
    geom_point(shape = 21) + 
    facet_wrap(~key)

immunogenomics/LISI documentation built on July 30, 2020, 5:49 a.m.