NDCN RNAseq compbio office hour
When you get your high-throughput sequencing data from the sequencing facility, it usualy comes in the FASTQ format. This is a file that has four lines for each sequence:
- A line that starts with the
@symbol that is followed by a sequence identifier. - The sequence (typically A, C, T, G or N).
- A line that starts with the
+symbol, and can sometimes have the sequence identifier from line 1. - The quality values for the sequence in line 2.
You can also obtain FASTQ files from Gene Expression Omnibus and Short Read Archive, where many people have deposited published datasets, or from public consortia such as ENCODE.
Whether you are generating your own sequences, or downloading them from a resource or collaborator, it is always a good idea to do some quick checks on the FASTQ files to see if there are any issues that might cause issues with analyses downstream. Here, we use FastQC to run some basic diagnostics on some FASTQ files.
From FastQC website: "FastQC aims to provide a simple way to do some quality control checks on raw sequence data coming from high throughput sequencing pipelines. It provides a modular set of analyses which you can use to give a quick impression of whether your data has any problems of which you should be aware before doing any further analysis. "
Unless you are processing the raw data yourself, most of the time you would receive the processed data from your bioinformatician/ computational analyst. For RNA-seq, this is typically a table of gene expression counts for each of the libraries that you sequenced.
It is always a good idea to take a look at that processed data yourself, and see if there are things in the data that are out of the ordinary, and might cause problem further downstream.
It is always recommended to have replicates for RNAseq experiments, since it allows differential analysis programs to model the intrinsic biological variations of your samples and only extract differences that are related to the question that you're trying to answer.
However, "variation" in your samples can also come from "technical" sources that are not part of the biological question (e.g. culture conditions/reagents, transfection efficiency, library prep). One way to see its impact is to have biological replicates, and look for similarities of experimental conditions between the replicates. Sample correlation is a simple but useful way to quickly assess this.
Another approach to look at variation in your dataset is to perform a principal component analysis (PCA). It tries to reduce the number of variables in your dataset into a small set of features (components) that could still describe the variation in the data. It also enables visualization of the "spread" of samples based on various principal components.
The most common question that we have when doing RNA-seq is looking for gene expression differences between two criteria (e.g. experiment vs control, patients vs unaffected). One way to do this is to run a differential expression analysis.
There are numerous methods that have been developed, and the one used
in this exercise has been incorporated to many bulk RNA-seq
pipelines: the DESeq2 R package.
Once the differential analysis is done, you can do some simple visualizations to take a look at the data, and see if it makes sense.