Difference between revisions of "Guided linear model analysis"

Revision as of 18:15, 9 December 2020

Analysis title

Guided linear model analysis

Provider

geneXplain GmbH

Class

DiffExpressDesigner

Plugin

com.genexplain.stat (geneXplain Stat)

Linear model analysis using Limma with experimental design specified through an annotation table

Description

This tool performs linear model analysis on the given input table guided by selected experimental factors defined in a sample table. The analysis aims at finding significant differences between pairs of levels of a main factor. Furthermore, an ANOVA is carried out for all contrasts together. The assignment of main factor levels to columns of the input table is specified in a column of a sample table. Additional variables can be controlled for by providing their column names in the sample table. Moreover, Surrogate Variable Analysis can be included to infer unspecified factors.

Please consider that the column names of the input table must correspond to names of rows in the sample table. If in the sample table the input column names are not the IDs one can specify the sample table column that contains the correct names (Sample column). If only a subset of input table columns shall be used this can be specified in the Data columns parameter.

Further to avoid possible issue related to the input format, ensure that column/sample names cannot be confused with numbers. Table column names should be compatible with R.

Also, please ensure that Input type and Normalization methods are correct. Raw counts will be processed using Limma's voom method, optionally including the specified normalization method, whereas Normalized expression values are used as is, and for Transformed counts an intensity-based trend is included during Limma analysis (eBayes parameter trend=TRUE)

Parameters

• Input table - Path to table with input data
• Input type - Specify type of input data
• Normalization method - Normalization to apply with voom
• Data columns - Optionally specify a subset of input columns
• Sample table - Table with sample (column) annotation
• Sample column - Annotation table column that contains data sample names, if row names are not samples
• Main factor - Main factor to define comparisons
• Reference level - Optional reference/base level. This level will be subtracted from other levels to form contrasts
• Compare to reference only - Include in contrasts only comparisons between the reference and other levels
• Control factors - Optionally specify a set of sample columns as control factors
• With SVA - Use Surrogate Variable Analysis to account for unobserved factors
• Robust - Use procedures robustified against outlier sample variances
• Output folder - Folder for output items

Output

The output is stored in the specified folder and contains one result table for each contrast, one ANOVA table for all coefficients as well as the resulting design matrix that shows the assignment of input columns to factor levels. If the main factor has only two levels the ANOVA table is equivalent to the single contrast result table that is produced by this analysis.

The output contains the columns described below. Columns highlighted in bold are shown in the default view. The other columns can be included on demand via the Columns tab of the lower right panel (available with opened output table).

Contrast result table

logFC

Fold change (log)

CI.025

Fold change (Lower confidence interval)

CI.975

Fold change (Upper confidence interval)

AveExpr

Average log2-expression for the probe over all arrays

t

Moderated T-statistic

P.Value

P-value Differential expression

adj.P.Val

Adjusted P-value (Benjamini-Hochberg)

B

Log-odds that the gene / probe shows differential expression

L10pval

Modulus of the decadic P-value logarithm

L10fdr

Modulus of the decadic adjusted P-value logarithm

signL10pval

Decadic P-value logarithm with same sign as the log fold change

signL10fdr

Decadic adjusted P-value logarithm with same sign as the log fold change

ANOVA result table

In an ANOVA table for more than two main factor levels, the first columns are the contrasts deduced from the main factor. The other columns are as follows. Further information is provided by the Limma userguide.

AveExpr

Average expression

F

F statistic

P.Value

F test P-value

adj.P.Val

Adjusted P-value

L10pval

Modulus of the decadic P-value logarithm

L10fdr/

Modulus of the decadic adjusted P-value logarithm

Reference

Smyth, G. K. (2005). Limma: linear models for microarray data. In: Bioinformatics and Computational Biology Solutions using R and Bioconductor. R. Gentleman, V. Carey, S. Dudoit, R. Irizarry, W. Huber (eds), Springer, New York, 2005.

limma: Linear Models for Microarray and RNA-Seq Data User’s Guide

Leek, J.T., Johnson, W.E., Parker, H.S., Jaffe, A.E., and Storey J.D. (2012) The sva package for removing batch effects and other unwanted variation in high-throughput experiments. Bioinformatics 28:882–883.