In fboulnois/made: Microarray Analysis of Differential Expression

# Set global report options
knitr::read_chunk(system.file("extdata", "report-code.R", package = "made"))
knitr::opts_chunk$set(echo = FALSE, fig.width = 12, fig.height = 8)
options(width = 120)

Introduction

This report summarizes the results of an oligonucleotide microarray experiment using the MADE pipeline. To cite the MADE package in publications type citation("made") in R.

The analysis involves the following steps:

1. Quality Assessment, to identify common errors and artifacts with the experiment.
2. Normalization, which adjusts the data such that meaningful biological comparisons can be identified.
3. Differential expression, which compares gene expression across groups defined by the user.
4. Coexpression, which identifies correlation patterns in gene expression.
5. Pathway Analysis, which determines which biological terms and pathways are associated to the experiment.

Experimental setup

The following sections summarize the options and experimental setup of the microarray pipeline.

User options

Group and sample information

A total of r length(sampleInfo$samples) samples were analyzed across r length(sampleInfo$groups) different groups. The groups were compared as follows: r config$groups$contrast_groups.

Quality Assessment

The following sections include important indicators relating to the overall quality of the microarray experiment. The log-intensity values distribution identifies the quality of the samples, while the hierarchical dendrogram identifies the quality of the groups. Sufficient heterogeneity in either of these indicators may reveal an underlying problem with the experiment.

Log-intensity values distribution

The sample quality can be summarized in a log-intensity values distribution plot. The x-axis represents the probe intensity in log2 scale and the y-axis represents the probe density.r if(hasBatches){ " Samples are colored by which scan dates they belong to." } The resulting intensity distribution for each sample is plotted. In the ideal case, samples should have similar intensity profiles. If a sample has a distribution which is very different from the other distributions, this typically indicates an outlier. If multiple groups of distributions are observed instead then this may represent an unknown source of variation which can confound the variables of interest.

Hierarchical clustering of samples

The quality of each group can be summarized in a dendrogram of the hierarchical clustering of samples. The Spearman distance metric is calculated to assess the similarity between each sample. The vertical height between each node indicates how closely correlated the samples are. Samples are colored by which group they belong to. Samples which are more closely correlated to each other are clustered together. Samples which are far from all other samples may represent outliers. In the ideal case, all samples from the same group should be tightly clustered together as homogeneous bands of color. If the groups are mixed up such that unrelated samples are clustered together than this may represent an unknown source of variation which can confound the variables of interest.

Normalization

Differential Expression

The top 50 most differentially expressed genes for each group comparison are as follows:

Coexpression

Coexpression identifies correlated patterns in gene expression across samples. This is useful to find groups of genes which are expressed together either positively or negatively. A value between -1 and 1 is obtained for each pair of genes using the Spearman distance metric and then transformed into a color value: bright red indicates that the genes are perfectly positively correlated (i.e., they coexpress positively or negatively across samples), dark blue indicates that the genes are perfectly negatively correlated (i.e. when one is overexpressed the other is underexpressed or vice versa), and white indicates that the genes are not correlated. A gene always has perfect positive correlation with itself (i.e. the diagonal on the heatmap). The top 50 most correlated genes for each group comparison are as follows:

Pathway analysis

The following sections represent biological terms or pathways that are statistically associated to the differentially expressed genes in each comparison of interest. The analyses are similar to using the hypergeometric distribution to find the probability of observing n or more differentially expressed genes annotated with a specific term amongst all genes available on the microarray.

Gene Ontology (GO) term analysis

The Gene Ontology (GO) term analysis identifies common biological terms among the set of differentially expressed genes. The Gene Ontology uses a hierarchical representation of biological terms, where each term may have zero or more child terms. The hierarchy is composed of three top-level terms: Biological Process (BP), Cellular Component (CC), and Molecular Function (MF). Gene identifiers may be associated to terms or child terms. Because genes may be associated to multiple related terms in the GO hierarchy, during the analysis there is an additional step known as "conditioning" which successively prunes more general terms until only the most specific terms remain. The top 20 most significant GO terms for each top-level term in each group comparison are as follows:

Reactome pathway analysis

Reactome is an open-source, manually curated and peer reviewed database which links biological pathways with genes and proteins. The Reactome pathway analysis identifies common biological pathways among the set of differentially expressed genes. The top 20 most significant Reactome pathways for each group comparison are as follows:

Other information

fboulnois/made documentation built on May 16, 2019, 12:01 p.m.