README.md

In smtnkc/sliced: Constructs very large correlation matrices

sliced

R package to construct and export very large correlation matrices.

Installation

1. First, install devtools on R console by: install.packages("devtools")
2. Then, install sliced package by: devtools::install_github("smtnkc/sliced")
3. Note that, sliced package depends on ff which can be installed by: install.packages("ff")

Usage

An example R file that is using the sliced package should look like:

library(readr)
library(sliced)

DATA_FRAME  <- read_csv("your_large_dataset.csv")
N_BLOCKS    <- 5  # Number of blocks you want to slice your dataframe
VERBOSE     <- TRUE 

# Get the first column as a vector
ROW_NAMES <- unlist(DATA_FRAME[,1])

# Transpose without the first column
TRANSPOSED_DF <- as.data.frame(t(DATA_FRAME[,-1]))

# Set colnames as rownames
colnames(TRANSPOSED_DF) <- ROW_NAMES

# Construct the virtual correlation matrix
COR_MATRIX <- sliced.cor(TRANSPOSED_DF, N_BLOCKS, "pearson", VERBOSE)

colnames(COR_MATRIX) <- ROW_NAMES
rownames(COR_MATRIX) <- ROW_NAMES

# Export the matrix as a CSV file
COR_FILE <- "your_output_file.csv"
sliced.write(COR_MATRIX, COR_FILE, N_BLOCKS, VERBOSE)

Important Notes :exclamation::exclamation::exclamation:

• N_BLOCKS parameter do not have to be a divisor of the number of columns your dataframe has. The package deals with it by adding remainder columns to the last slice. :wink:
• Transposition in the example above may not always be necessary. However, keep in mind that, correlation is always performed column based. Since most of the gene expression datasets keeps genes in rows and samples in columns, transposition is needed to get correlation between genes.

smtnkc/sliced documentation built on May 26, 2019, 4:33 a.m.