chickn

R package {chickn} implements Chromatogram Hierarchical Compressive K-means with Nystrom approximation clustering approach. It allows processing a large collection of high-resolution mass spectrometry signals (chromatographic profiles) relying on the compressed data version (data sketch). The Filebacked Big Matrix (FBM) class from the bigstatsr package is used to store and to manipulate matrices, which cannot be load in memory.

Pre-requisites

R version >= 3.5, bigstatsr >= 1.2.3

OS tested: Ubuntu

chickn R package dependencies: foreach, doRNG, parallel, doParallel, Rcpp, RcppArmadillo, RcppParallel, nloptr, pracma, stats, MASS,zipfR, mvnfast, rmio, Rdpack

Installation

You can install {chickn} from GitLab with:

devtools::install_gitlab(repo = "Olga.Permiakova/chickn")

Typical installation error

If the Rdpack package instalation had non-zero exit status, you should first install the development version of Rdpack from Github using:

devtools::install_github("GeoBosh/Rdpack")

then proceed with the {chickn} package installation.

Example

This is a basic example which demonstrates how the complete {chickn} pipeline can be applied to a small part of the UPS2 data (toy dataset provided in the package):

library(chickn)
data(UPS2)
n = nrow(UPS2)
N = ncol(UPS2)
## Convert data into FBM format, a .bk file is generated in the working directory
UPS2_FBM = bigstatsr::FBM(init = UPS2, ncol=N, nrow = n, backingfile = "UPS2_FBM")$save()
## Apply main chickn function. 
output  <- CHICKN_W1(Data = UPS2_FBM, K = 2, k_total =32, 
                     max_neighbors = 10,DIR_output = getwd(), DIR_tmp = tempfile(),
                     ncores = 4, N0 = N, DoPreimage = TRUE)
## NB: The generated output files (NystromKernel, Centroids_out, Cluster_assign_out) are stored in DIR_output directory.                      

Citing Our work

The published article of the project will be linked here.