R Tutorial: Differential expression analysis

Key Takeaways

welcome to differential expression analysis with lima i'm john blitzchalk and i'll be your instructor for this course you will learn how to analyze data generated by functional genomics experiments so let's review the terminology i will use throughout the course imagine a hypothetical experiment in which the researcher has treated cells with two different drugs a and b the drug treatment is the variable of an interest and is an example of a phenotype each sample is processed by a high throughput assay to measure thousands of genes these are the features in this course the features will always be genes but in other experiments they could be proteins or some other molecular feature of a cell for each feature the assay produces a value that is a proxy for the relative abundance of that feature for genes this is the number of rna transcripts that are expressed thus i will refer to the measurements as expression levels the statistical models you will build in this course will test for differences in these measurements between samples with different phenotypes if a feature has a higher expression level for one group relative to the other this is called upregulated conversely a lower expression level is called down regulated the overall goal is to identify the genes that are associated with a phenotype of interest some examples of differential expression studies would be identifying all the genes associated with a response to a stimulus like a drug a developmental process or a genetic mutation now why bother testing thousands of genes known as a genome-wide differential expression analysis when it would be easier to focus on measuring only a handful of genes the first motivation is novelty you may discover that some additional genes play an unexpected role in the process you are studying second is context interpreting the relevance of any one gene is easier when you can compare it to the behavior of all other genes third is to gain a systems level understanding of the process by measuring all genes simultaneously you can identify higher order systems like pathways that you would miss with a more targeted approach there are many steps to complete an experiment from designing the initial study to compiling the results in this course the main focus will be on creating linear models to test scientific hypotheses but you will also learn the basics of exploring the data and interpreting the test results thus in general this course is agnostic to the type of data collected to learn more about how to process the data generated by a specific technique check out the other bio conductor courses on data camp now for some caveats to keep in mind first the measurements recorded by these genomic technologies are relative not absolute thus comparing the measurements across samples is valid but it's not possible to directly convert these measurements to the total number of molecules second study design is very important all experimental techniques introduce some technical noise but the statistical methods for removing this noise are only valid for properly designed studies you'll learn about proper study design later in the course now let's test your comprehensive