Package ubic.gemma.core.analysis.expression.diff

Class LinearModelAnalyzer

java.lang.Object

ubic.gemma.core.analysis.expression.diff.LinearModelAnalyzer

All Implemented Interfaces:

DiffExAnalyzer

@Component
@ParametersAreNonnullByDefault
public class LinearModelAnalyzer
extends Object
implements DiffExAnalyzer

Handles fitting linear models with continuous or fixed-level covariates. Data are always log-transformed.

Interactions can be included if a DifferentialExpressionAnalysisConfig is passed as an argument to 'run'. Currently we only support interactions if there are two factors in the model (no more).

One factor can be constant (the same value for all samples); such a factor will be analyzed by looking at the intercept in the fitted model. This is only appropriate for 'non-reference' designs on ratiometric arrays.

This also supports subsetting the data based on a factor. For example, a data set with "tissue" as a factor could be analyzed per-tissue rather than with tissue as a covariate.

This only handles the analysis, not the persistence or output of the results.

Author:

paul

Constructor Summary

Constructors

Constructor	Description
LinearModelAnalyzer()

Method Summary

Modifier and Type	Method	Description
Collection<DifferentialExpressionAnalysis>	run(ExpressionExperiment ee, Map<FactorValue,ExpressionExperimentSubSet> subsets, ExpressionDataDoubleMatrix dmatrix, DifferentialExpressionAnalysisConfig config)	Analyze a dataset with a pre-existing subset structure.
Collection<DifferentialExpressionAnalysis>	run(ExpressionExperiment expressionExperiment, ExpressionDataDoubleMatrix dmatrix, DifferentialExpressionAnalysisConfig config)	I apologize for this being so complicated.
DifferentialExpressionAnalysis	run(ExpressionExperimentSubSet subset, ExpressionDataDoubleMatrix dmatrix, DifferentialExpressionAnalysisConfig config)	Analyze a subset.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Details

LinearModelAnalyzer

public LinearModelAnalyzer()

Method Details

run

public Collection<DifferentialExpressionAnalysis> run(ExpressionExperiment expressionExperiment,
 ExpressionDataDoubleMatrix dmatrix,
 DifferentialExpressionAnalysisConfig config)
                                               throws AnalysisException

I apologize for this being so complicated. Basically there are four phases:

1. Get the data matrix and factors
2. Determine baseline groups; build model and contrasts
3. Run the analysis
4. Postprocess the analysis

By far the most complex is #2 -- it depends on which factors and what kind they are.

Specified by:

run in interface DiffExAnalyzer

Parameters:

expressionExperiment - experiment to analyze

dmatrix - D matrix

config - config

Returns:

analyses. There will be more than one if a subset factor is defined.

Throws:

AnalysisException

run

public Collection<DifferentialExpressionAnalysis> run(ExpressionExperiment ee,
 Map<FactorValue,ExpressionExperimentSubSet> subsets,
 ExpressionDataDoubleMatrix dmatrix,
 DifferentialExpressionAnalysisConfig config)
                                               throws AnalysisException

Description copied from interface: DiffExAnalyzer

Analyze a dataset with a pre-existing subset structure.

A subset must be defined in the configuration.

Specified by:

run in interface DiffExAnalyzer

Throws:

AnalysisException

run

public DifferentialExpressionAnalysis run(ExpressionExperimentSubSet subset,
 ExpressionDataDoubleMatrix dmatrix,
 DifferentialExpressionAnalysisConfig config)
                                   throws AnalysisException

Description copied from interface: DiffExAnalyzer

Analyze a subset.

Note that normally when we run a subset analysis, the subsetting is done internally, so we pass in the expression experiment, not the subset. This method is used for exceptions to that.

Specified by:

run in interface DiffExAnalyzer

Parameters:

subset - subset

config - config

Returns:

analysis

Throws:

AnalysisException