Thursday Afternoon — Trimming Sequences

Credit: this material has been adapted from the Data Carpentry lesson Data Wrangling and Processing for Genomics. The adaptations principally include making sure commands work on the CLIMB-BIG-DATA Notebook servers. All Carpentries instructional material, and therefore all material here, is made available under the Creative Commons Attribution license. See https://github.com/datacarpentry/wrangling-genomics/blob/main/LICENSE.md.

Trimming Sequences

In the previous session, we took a high-level look at the quality of each of our samples using FastQC. We visualized per-base quality graphs showing the distribution of read quality at each base across all reads in a sample and extracted information about which samples fail which quality checks. Some of our samples failed quite a few quality metrics used by FastQC. This does not mean, though, that our samples should be thrown out! It is very common to have some quality metrics fail, and this may or may not be a problem for your downstream application.

For our variant-calling workflow, we will be removing some of the low quality sequences to reduce our false positive rate due to sequencing error.

We will use a program called Trimmomatic to filter poor quality reads and trim poor quality bases from our samples.

You should have the hang of installing propgrams with Comda by now. Get on and install Trimmomatic!

Trimmomatic options

Trimmomatic has a variety of options to trim your reads. If we run the following command, we can see some of our options.

$ trimmomatic

Which will give you the following output:

Usage: 
       PE [-version] [-threads <threads>] [-phred33|-phred64] [-trimlog <trimLogFile>] [-summary <statsSummaryFile>] [-quiet] [-validatePairs] [-basein <inputBase> | <inputFile1> <inputFile2>] [-baseout <outputBase> | <outputFile1P> <outputFile1U> <outputFile2P> <outputFile2U>] <trimmer1>...
   or: 
       SE [-version] [-threads <threads>] [-phred33|-phred64] [-trimlog <trimLogFile>] [-summary <statsSummaryFile>] [-quiet] <inputFile> <outputFile> <trimmer1>...
   or: 
       -version

This output shows us that we must first specify whether we have paired end (PE) or single end (SE) reads.

Next, we specify what flag we would like to run. For example, you can specify threads to indicate the number of processors on your computer that you want Trimmomatic to use. In most cases using multiple threads (processors) can help to run the trimming faster. These flags are not necessary, but they can give you more control over the command. The flags are followed by positional arguments, meaning the order in which you specify them is important.

In paired end mode, Trimmomatic expects the two input files, and then the names of the output files. These files are described below. While, in single end mode, Trimmomatic will expect 1 file as input, after which you can enter the optional settings and lastly the name of the output file.

The last thing Trimmomatic expects to see is the trimming parameters:

We will use only a few of these options and trimming steps in our analysis. It is important to understand the steps you are using to clean your data. For more information about the Trimmomatic arguments and options, see the Trimmomatic manual.

However, a complete command for Trimmomatic will look something like the command below. This command is an example and will not work, as we do not have the files it refers to:

$ trimmomatic PE -threads 4 SRR_1056_1.fastq SRR_1056_2.fastq  \
              SRR_1056_1.trimmed.fastq SRR_1056_1un.trimmed.fastq \
              SRR_1056_2.trimmed.fastq SRR_1056_2un.trimmed.fastq \
              ILLUMINACLIP:SRR_adapters.fa SLIDINGWINDOW:4:20

In this example, we have told Trimmomatic:

Multi-line commands

Some of the commands we ran in this lesson are long! When typing a long command into your terminal, you can use the \ character to separate code chunks onto separate lines. This can make your code more readable.

Running Trimmomatic

Now we will run Trimmomatic on our data. To begin, navigate to your untrimmed_fastq data directory:

$ cd ~/dc_workshop/data/untrimmed_fastq

We are going to run Trimmomatic on one of our paired-end samples. While using FastQC we saw that Nextera adapters were present in our samples. The adapter sequences came with the installation of trimmomatic, so we will first copy these sequences into our current directory.

$ cp /home/jovyan/.conda/pkgs/trimmomatic-0.39-hdfd78af_2/share/trimmomatic-0.39-2/adapters/NexteraPE-PE.fa .

We will also use a sliding window of size 4 that will remove bases if their phred score is below 20 (like in our example above). We will also discard any reads that do not have at least 25 bases remaining after this trimming step. Three additional pieces of code are also added to the end of the ILLUMINACLIP step. These three additional numbers (2:40:15) tell Trimmimatic how to handle sequence matches to the Nextera adapters. A detailed explanation of how they work is advanced for this particular lesson. For now we will use these numbers as a default and recognize they are needed to for Trimmomatic to run properly. This command will take a few minutes to run.

$ trimmomatic PE SRR2589044_1.fastq.gz SRR2589044_2.fastq.gz \
                SRR2589044_1.trim.fastq.gz SRR2589044_1un.trim.fastq.gz \
                SRR2589044_2.trim.fastq.gz SRR2589044_2un.trim.fastq.gz \
                SLIDINGWINDOW:4:20 MINLEN:25 ILLUMINACLIP:NexteraPE-PE.fa:2:40:15

TrimmomaticPE: Started with arguments:
 SRR2589044_1.fastq.gz SRR2589044_2.fastq.gz SRR2589044_1.trim.fastq.gz SRR2589044_1un.trim.fastq.gz SRR2589044_2.trim.fastq.gz SRR2589044_2un.trim.fastq.gz SLIDINGWINDOW:4:20 MINLEN:25 ILLUMINACLIP:NexteraPE-PE.fa:2:40:15
Multiple cores found: Using 2 threads
Using PrefixPair: 'AGATGTGTATAAGAGACAG' and 'AGATGTGTATAAGAGACAG'
Using Long Clipping Sequence: 'GTCTCGTGGGCTCGGAGATGTGTATAAGAGACAG'
Using Long Clipping Sequence: 'TCGTCGGCAGCGTCAGATGTGTATAAGAGACAG'
Using Long Clipping Sequence: 'CTGTCTCTTATACACATCTCCGAGCCCACGAGAC'
Using Long Clipping Sequence: 'CTGTCTCTTATACACATCTGACGCTGCCGACGA'
ILLUMINACLIP: Using 1 prefix pairs, 4 forward/reverse sequences, 0 forward only sequences, 0 reverse only sequences
Quality encoding detected as phred33
Input Read Pairs: 1107090 Both Surviving: 885220 (79.96%) Forward Only Surviving: 216472 (19.55%) Reverse Only Surviving: 2850 (0.26%) Dropped: 2548 (0.23%)
TrimmomaticPE: Completed successfully

Quick questions:

Use the output from your Trimmomatic command to answer the following questions.

1) What percent of reads did we discard from our sample? 2) What percent of reads did we keep both pairs?

You may have noticed that Trimmomatic automatically detected the quality encoding of our sample. It is always a good idea to double-check this or to enter the quality encoding manually.

We can confirm that we have our output files:

$ ls SRR2589044*

SRR2589044_1.fastq.gz       SRR2589044_1un.trim.fastq.gz  SRR2589044_2.trim.fastq.gz
SRR2589044_1.trim.fastq.gz  SRR2589044_2.fastq.gz         SRR2589044_2un.trim.fastq.gz

The output files are also FASTQ files. It should be smaller than our input file, because we have removed reads. We can confirm this:

$ ls SRR2589044* -l -h

-rw-rw-r-- 1 dcuser dcuser 124M Jul  6 20:22 SRR2589044_1.fastq.gz
-rw-rw-r-- 1 dcuser dcuser  94M Jul  6 22:33 SRR2589044_1.trim.fastq.gz
-rw-rw-r-- 1 dcuser dcuser  18M Jul  6 22:33 SRR2589044_1un.trim.fastq.gz
-rw-rw-r-- 1 dcuser dcuser 128M Jul  6 20:24 SRR2589044_2.fastq.gz
-rw-rw-r-- 1 dcuser dcuser  91M Jul  6 22:33 SRR2589044_2.trim.fastq.gz
-rw-rw-r-- 1 dcuser dcuser 271K Jul  6 22:33 SRR2589044_2un.trim.fastq.gz

We have just successfully run Trimmomatic on one of our FASTQ files! However, there is some bad news. Trimmomatic can only operate on one sample at a time and we have more than one sample. The good news is that we can use a for loop to iterate through our sample files quickly!

We unzipped one of our files before to work with it, let’s compress it again before we run our for loop.

gzip SRR2584863_1.fastq 

$ for infile in *_1.fastq.gz
do
   base=$(basename ${infile} _1.fastq.gz)
   trimmomatic PE ${infile} ${base}_2.fastq.gz \
                ${base}_1.trim.fastq.gz ${base}_1un.trim.fastq.gz \
                ${base}_2.trim.fastq.gz ${base}_2un.trim.fastq.gz \
                SLIDINGWINDOW:4:20 MINLEN:25 ILLUMINACLIP:NexteraPE-PE.fa:2:40:15 
done

Go ahead and run the for loop. It should take a few minutes for Trimmomatic to run for each of our six input files. Once it is done running, take a look at your directory contents. You will notice that even though we ran Trimmomatic on file SRR2589044 before running the for loop, there is only one set of files for it. Because we matched the ending _1.fastq.gz, we re-ran Trimmomatic on this file, overwriting our first results. That is ok, but it is good to be aware that it happened. Unix commands often overwrite existing files without telling you.

$ ls

NexteraPE-PE.fa               SRR2584866_1.fastq.gz         SRR2589044_1.trim.fastq.gz
SRR2584863_1.fastq.gz         SRR2584866_1.trim.fastq.gz    SRR2589044_1un.trim.fastq.gz
SRR2584863_1.trim.fastq.gz    SRR2584866_1un.trim.fastq.gz  SRR2589044_2.fastq.gz
SRR2584863_1un.trim.fastq.gz  SRR2584866_2.fastq.gz         SRR2589044_2.trim.fastq.gz
SRR2584863_2.fastq.gz         SRR2584866_2.trim.fastq.gz    SRR2589044_2un.trim.fastq.gz
SRR2584863_2.trim.fastq.gz    SRR2584866_2un.trim.fastq.gz
SRR2584863_2un.trim.fastq.gz  SRR2589044_1.fastq.gz

Quick question:

• We trimmed our fastq files with Nextera adapters, but there are other adapters that are commonly used.
• What other adapter files came with Trimmomatic?
• Clue: Take a look in /home/jovyan/.conda/pkgs/trimmomatic-0.39-hdfd78af_2/share/trimmomatic-0.39-2/adapters/

We have now completed the trimming and filtering steps of our quality control process! Before we move on, let’s move our trimmed FASTQ files to a new subdirectory within our data/ directory.

$ cd ~/dc_workshop/data/untrimmed_fastq
$ mkdir ../trimmed_fastq
$ mv *.trim* ../trimmed_fastq
$ cd ../trimmed_fastq
$ ls

SRR2584863_1.trim.fastq.gz    SRR2584866_1.trim.fastq.gz    SRR2589044_1.trim.fastq.gz
SRR2584863_1un.trim.fastq.gz  SRR2584866_1un.trim.fastq.gz  SRR2589044_1un.trim.fastq.gz
SRR2584863_2.trim.fastq.gz    SRR2584866_2.trim.fastq.gz    SRR2589044_2.trim.fastq.gz
SRR2584863_2un.trim.fastq.gz  SRR2584866_2un.trim.fastq.gz  SRR2589044_2un.trim.fastq.gz

This kind of good housekeeping in terms of storing files in clearly identified and structured directories is essential for sound bioinformatics research. You might think you know what everything is on the day you do the analysis but you can guarantee you won’t be able to remember when you return from holiday or want to share an analysis with someone else.

And before we finish for the day

Now that our samples have gone through quality control, they should perform better on the quality tests run by FastQC. Go ahead and re-run FastQC on your trimmed FASTQ files and visualize the HTML files to see whether your per base sequence quality is higher after trimming.

After trimming and filtering, what we hope is that our overall quality is much higher, we have a distribution of sequence lengths, and more samples pass adapter content. However, quality trimming is not perfect, and some programs are better at removing some sequences than others. Because our sequences still contain 3’ adapters, it could be important to explore other trimming tools like cutadapt to remove these, depending on your downstream application. Trimmomatic did pretty well though, and its performance is good enough for our workflow.

End of Today’s session

And that’s it for today. See you on Friday!