So in this video, we are going to to discuss one of the most important data containers in Bioconductor. An old data container called an ExpressionSet. As the name implies, this is a data container for representing an expression experiment. This is a type of experiment where we're using either microwaves or RNA sequencing, the expression of perhaps thousands of genes on hundreds of samples. So in an expression set, the first component we have is an expression matrix. This is a matrix of actual measured expression values. In Bioconductor we follow the same convention that you do in most of computational biology, where these type of matrices have samples on the columns and features on the rows. Features are genes or probes. And this matrix is a single matrix of numeric values. Each entry into the matrix tells you something about how that particular gene or feature was expressed in that particular sample. Together with the expression data we need additional information or metadata on the samples. So that is represented in context of a different matrix. Or a data frame in R where each row of this matrix corresponds to a sample and each column corresponds to a different covariant. So examples of covariants could be sample six, it could be sample eight, it could be time of the experiment, or date of the experiment. And other information on the individual samples. These two matrices, or a matrix and a data frame are linked together in the sense that one row of the phenotype data corresponds to one column of the expression matrix. We also have metadata on the genes or on the features we are missing. This is information about what feature is it actually we are missing, right? So for example, on a micro rain in a gene expression experiment, we are not really measuring a specific gene. We are measuring a specific RNA sequence that represents a gene. So a type of metadata on the feature could be what sequence was actually measured and what gene does this sequence correspond to. In the same way as the phenoData and the experiment data are linked together, the feature data and the experiment data are linked together in the sense. One row of the feature data correspond to one row of the expression data. So in context of the previous lecture, we think of the expression matrix as really raw experimental data. We think of the phenotype data and the feature data as meta data on the experiment. And all of these three data frames or matrixes fit together into something we call an ExpressionSet, which provide a tied linking between rows of the matrix, rows of the expression matrix, columns of the expression matrix, and the relevant rows and columns on the feature data and the phenotype data. This is a very, very common type of construction in Bioconductor that has proven to be immensely powerful. There's an obvious generalization the expression set into something called an eSet. And the difference between an expression set and an eSet is that we don't have single expression matrix. We have multiple matrices. So what could that correspond to? For example, it could correspond to a two column micro experiment where each matrix is a column of the microarray. Or it could correspond to data on say DNA methylation, where one matrix corresponds to the methylation channel and the other matrix correspond to the un-methylation channel.
