OOMPA Overview

OOMPA PROJECT

Class Prediction

The Class Prediicton R-Forge project is one component of the broader OOMPA Project produced by the Coombes Lab over the past decade. it consists of a set of packages to help select features and build models in order to predict binary outcomes from omics data sets.

List of Packages

BimodalIndex

The Bimodality Index is a tool developed in a paper by Wang and colleagues to identify features with bimodal distributions. The motivating idea is that bimodally expressed features are always markers of something; hopefully, a collection of those somethings will be related to the outcome to be predicted. See also this paper by Tong and colleagues and the SIBER package for an extension of this idea to RNA-Sequencing data.

Modeler

Classes and Methods for Training and Using Binary Prediction Models. Defines classes and methods for training and using binary prediction models. it implements a variety of methods for filtering features (i.e., removing features that you are sure won't help), feature selection, and model construction. These are all provided with a common interface that allows you to use them in other contexts such as cross validation.

CrossValidate

Classes and Methods for Cross Validation of "Class Prediction" Algorithms Defines classes and methods to cross-validate various binary classification algorithms used for "class prediction" problems. These work with Modeler objects, but can make sure that feature filtering or feature selection are included properly within the cross-validation loop.

GenAlgo

Classes and Methods to Use Genetic Algorithms for Feature Selection Defines classes and methods that can be used to implement genetic algorithms for feature selection. The idea is that we want to select a fixed number of features to combine into a linear classifier that can predict a binary outcome, and can use a genetic algorithm heuristically to select an optimal set of features.

ClassPrediction

Classes and Methods for "Class Prediction" with Microarrays or Proteomics Bundles the classes and pacakges used to help with "class prediction" problems in the OOMPA project. Class prediction consists of supervised methods to select relevant features and construct predictive models for binary, ordinal, continuous, or survival outcomes. The ClassPrediction package provides no new functionality. The only thing it does is automatically load the GenAlgo and CrosValidate packages for concurrent usage. Never submitted to CRAN.

Condens8R

The Condens8R package provides no new functionality. The only thing it does is automatically load the GenAlgo and CrosValidate packages for concurrent usage. Under Development.