Workflow design

Objects

ExpressionSet

ExpressoinSet introduction by Seth Falcon et al.

Dissecting an ExpressionSet object http://telliott99.blogspot.cz/2011/08/dissecting-expressionset-object.html

abbrevs: features F, samples S, covariates V,

contains:

• assayData (gene expression matrix FxS)
• phenotypeData (qualitative and quantitative data SxV) + metadata
  • also contains "class" named as "User_class"
• annotation (e.g., hgu95v2)
• featureData
• experimentData (minimum information about microarray experiment)
• protocolData

methods:

• get feature names
• get sample names
• access phenotype (e.g., gender)
• subsetting
• subset by phenotype (e.g., only male samples)
• transform into R object note: the minimal ExpressionSet object can be initialized by a gene expression matrix only.

Dependent variable (class)

(obsolete)

• type: qualitative, quantitative
• 1xS

PlatformAnnotation

• allows conversion into a collection of WU sets

Containers

• training and testing pair (training ExpressionSet, testing ExpressionSet)
  • also includes dependent variable definition
• set (e.g. set of traning and tesing pairs)
• ranking
  • note that some methods return only "limited" ranking (e.g. only the first top 200 from 4000 possible)
• wu set
  • head type, head description (e.g., URL), tail type, tail members
  • in future, working units able represent "inhibition" or "activation"

Plugins

The plugin description is here.

• load gene expression data
  • collapse into gene symbols or ENTREZ IDs
  • normalize data
  • show boxplot
  • accepts:
  • returns: ExpressionSet
• impute/remove missing values
  • methods
  • accepts: ExpressionSet (also in pair container)
  • returns: transformed ExpressionSet (also in pair container)
• apply set aggregation
  • mean, median, pca, setsig (generally arguments can be: ES_tr, ES_tt, C_tr, C_tt)
  • accepts: annotation, wu set, ExpressionSet (also in pair container)
  • returns: transformed ExpressionSet (also in pair container)
• apply feature selection
  • SVMRFE (select n param), infgain, t-test
  • accepts: ExpressionSet (also the first member from the pair container)
  • returns: ranking
• apply threshold
  • %, n of selected features
  • accepts: ranking , ExpressionSet (also in pair container)
  • returns: transformed ExpressionSet (also in pair container)
• cross validation (meta plugin)
  • set # of folds, stratification, random seed generator
  • accepts: dependent variable description , ExpressionSet (also the first member from the pair container)
  • returns: set of results
• visualization
  • PCA
    • accepts: dependent variable description , ExpressionSet (also the first member from the pair container)
  • model
    • accepts: model
  • CV results
    • accepts: statistic on set of CV numeric results
• ML algorithms
  • number of folds in case of just one input dataset
  • accepts: dependent variable description , ExpressionSet (also in pair container)
  • returns: model, acc, AUC, ...
• Statistical test
  • accepts: dependent variable description , ExpressionSet (also the first member from the pair container)
  • returns: ranking, result table

Design tool visualization

Example