Welcome to the blog

Posts

My thoughts and ideas

Pre-alignment QC | Griffith Lab

RNA-seq Bioinformatics

Introduction to bioinformatics for RNA sequence analysis

Pre-alignment QC


RNA-seq_Flowchart


You can use FastQC to get a sense of your data quality before alignment:

Video Tutorial here:

Try to run FastQC on your fastq files:

    cd $RNA_HOME/data
    fastqc *.fastq.gz

Then, go to the following url in your browser:

  • http://YOUR_DNS_NAME/workspace/rnaseq/data/
  • Note, you must replace YOUR_DNS_NAME with your own amazon instance DNS (e.g., ec2-54-187-159-113.us-west-2.compute.amazonaws.com))
  • Click on any of the *_fastqc.html files to view the FastQC report

Exercise: Investigate the source/explanation for over-represented sequences:

  • HINT: Try BLASTing them.

PRACTICAL EXERCISE 4

Assignment: Run FASTQC on one of the additional fastq files you downloaded in the previous practical exercise.

  • Hint: Remember that you stored this data in a separate working directory called ‘practice’.

Run FASTQC on the file ‘hcc1395_normal_1.fastq.gz’ and answer these questions by examining the output.

Questions

  • How many total sequences are there?
  • What is the range (x - y) of read lengths observed?
  • What is the most common average sequence quality score?
  • What does the Adaptor Content warning tell us?

Solution: When you are ready you can check your approach against the Solutions.


Run MultiQC on your fastqc reports to generate a single summary report across all samples/replicates.

    cd $RNA_HOME/data
    multiqc .
RNAseq Data | Griffith Lab

RNA-seq Bioinformatics

Introduction to bioinformatics for RNA sequence analysis

RNAseq Data


RNA-seq_Flowchart


Obtain RNA-seq test data.

The test data consists of two commercially available RNA samples: Universal Human Reference (UHR) and Human Brain Reference (HBR). The UHR is total RNA isolated from a diverse set of 10 cancer cell lines. The HBR is total RNA isolated from the brains of 23 Caucasians, male and female, of varying age but mostly 60-80 years old.

In addition, a spike-in control was used. Specifically we added an aliquot of the ERCC ExFold RNA Spike-In Control Mixes to each sample. The spike-in consists of 92 transcripts that are present in known concentrations across a wide abundance range (from very few copies to many copies). This range allows us to test the degree to which the RNA-seq assay (including all laboratory and analysis steps) accurately reflects the relative abundance of transcript species within a sample. There are two ‘mixes’ of these transcripts to allow an assessment of differential expression output between samples if you put one mix in each of your two comparisons. In our case, Mix1 was added to the UHR sample, and Mix2 was added to the HBR sample. We also have 3 complete experimental replicates for each sample. This allows us to assess the technical variability of our overall process of producing RNA-seq data in the lab.

For all libraries we prepared low-throughput (Set A) TruSeq Stranded Total RNA Sample Prep Kit libraries with Ribo-Zero Gold to remove both cytoplasmic and mitochondrial rRNA. Triplicate, indexed libraries were made starting with 100ng Agilent/Strategene Universal Human Reference total RNA and 100ng Ambion Human Brain Reference total RNA. The Universal Human Reference replicates received 2 ul of 1:1000 ERCC Mix 1. The Human Brain Reference replicates received 1:1000 ERCC Mix 2. The libraries were quantified with KAPA Library Quantification qPCR and adjusted to the appropriate concentration for sequencing. The triplicate, indexed libraries were then pooled prior to sequencing. Each pool of three replicate libraries were sequenced across 2 lanes of a HiSeq 2000 using paired-end sequence chemistry with 100bp read lengths.

So to summarize we have:

Each data set has a corresponding pair of FastQ files (read 1 and read 2 of paired end reads). The reads are paired-end 101-mers generated on an Illumina HiSeq instrument. The test data has been pre-filtered for reads that appear to map to chromosome 22. Lets copy the raw input data to our tutorial working directory.

    echo $RNA_DATA_DIR
    mkdir -p $RNA_DATA_DIR
    cd $RNA_DATA_DIR
    wget http://genomedata.org/rnaseq-tutorial/HBR_UHR_ERCC_ds_5pc.tar

Unpack the test data. You should see 6 sets of paired end fastq files. One for each of our sample replicates above. We have 6 pairs (12 files) because in fastq format, read 1 and read 2 of a each read pair (fragment) are stored in separate files.

    tar -xvf HBR_UHR_ERCC_ds_5pc.tar
    ls

Enter the data directory and view the first two read records of a file (in fastq format each read corresponds to 4 lines of data)

    zcat UHR_Rep1_ERCC-Mix1_Build37-ErccTranscripts-chr22.read1.fastq.gz | head -n 8

Identify the following components of each read: read name, read sequence, and quality string

How many reads are there in the first library? Decompress file on the fly with ‘zcat’, pipe into ‘grep’, search for the read name prefix and pipe into ‘wc’ to do a word count (‘-l’ gives lines)

    zcat UHR_Rep1_ERCC-Mix1_Build37-ErccTranscripts-chr22.read1.fastq.gz | grep -P "^\@HWI" | wc -l

PRACTICAL EXERCISE 3

Assignment: Download an additional dataset and unpack it. This data will be used in future practical exercises.

The practice dataset includes 3 replicates of data from the HCC1395 breast cancer cell line and 3 replicates of data from HCC1395BL matched lymphoblastoid line. So, this will be a tumor vs normal (cell line) comparison. The reads are paired-end 151-mers generated on an Illumina HiSeq instrument. The test data has been pre-filtered for reads that appear to map to chromosome 22.

Questions

Solution: When you are ready you can check your approach against the Solutions.


NOTE: various data sets used over time for our RNA-seq workshops can be found here: https://xfer.genome.wustl.edu/gxfer1/project/gms/testdata/bams/

If you use this data, please cite our paper: Citation