README.md

In smanchan96/binspec: Binary Classification on Mass Spectra

A preprocessing/peak identification, classification, and feature ranking R library for mass spectra. The preprocessing algorithm is an extension of Yasui et al.'s peak-finding algorithm (https://www.biostat.wisc.edu/~kbroman/teaching/statgen/2004/refs/yasui.pdf). The classifiers available are radial kernel support vector machines (tuned with leave-one-out classification) and random forests (tuned with out-of-bag error), with the random forests providing a utility for ranking features based on mean decrease in Gini impurity. We also provide a parallelized training function that allows for training classifiers on multiple sets of preprocessing parameters at once.

The package has uses the following libraries: * e1071 for support vector machines * randomForest for random forests * parallel for running tests with different parameter settings in parallel * roxygen2 and devtools for installing and documenting packages which can be installed in the R console using the command:

install.packages(c("e1071", "randomForest", "parallel", "roxygen2", "devtools"))

In order to install the package in R on a unix-based system, run the following commands after installing roxygen2 and

git clone https://github.com/smanchan96/binspec.git
cd binspec
./build.sh

If you are not on a unix-based system, you will not have access to the build script, so you will need to run build.R in the R console. Then you can get started in the R console by running

library(binspec)

Documentation is available here and will automatically be available in your R console after installation, using ?function-name. An rough example of us using this package is in example.R. Of course, it will not run without the mass spec data, but it is an okay template for figuring out how to use the packages.

smanchan96/binspec documentation built on May 30, 2019, 3:06 a.m.