Tag Archives: FASTQ

Data Management – Convert Oly PacBio H5 to FASTQ

After working with all of this Olympia oyster genome sequencing data, I remembered that we had an old, singular PacBio SMRT cell file (from June 2013). This file didn’t seem to be included in any recent assemblies of Sean’s or mine. This is most likely because we have it in the PacBio H5 format and not in FASTQ.

I installed PacBio’s pbh5tools on my computer (swoose), converted the file and moved it to owl/nightingales/O_lurida

python bash5tools.py /mnt/owl/nightingales/O_lurida/m130619_081336_42134_c100525122550000001823081109281326_s1_p0.bas.h5 --outType fastq 

I generated an MD5 checksum and appended to the checksums.md5 file in /owl/nightingales/O_lurida using the following command:

md5sum m130619_081336_42134_c100525122550000001823081109281326_s1_p0.fastq | awk '{print $2 " = " $1}' >> checksums.md5

The command above pipes the output to awk to format the output to match the existing format of the checksums.md5 file (i.e. filename = hash).

I’ve also updated our Nightingales spreadsheet (Google Sheet) to reflect this.

Will generate updated PacBio assemblies with Canu and/or Racon.

Data Management – Illumina Geoduck HiSeq & MiSeq Data

The HDD we received from Illumina last week only had data (i.e. fastq files) from the NovaSeq runs they performed – nothing from either the MiSeq, nor the HiSeq runs.

We contacted them about the missing data, they confirmed it was missing, and uploaded the remaining data to BaseSpace.

Began downloading the data – will take awhile…


Files will be temporarily stored in these locations:




Data Management – SRA Submission Olympia Oyster UW PacBio Data from 20170323

Submitted the FASTQ files from the UW PacBio Data from 20170323 to the NCBI sequence read archive (SRA).

FTP’d the data to NCBI’s servers, following their instructions. Briefly,

Change to the directory where the FASTQ files are (Owl/web/nightingales/O_lurida) and then initiate an FTP session:

ftp -i ftp-private.ncbi.nlm.nih.gov

Enter provided username/password, change to my designated uploads directory, create new folder dedicate to this particular upload. Then, upload all the files using the mput command:

mput *filtered_subreads*

SRA deets are below (assigned FASTQ files to existing BioProject and created a new BioSample). Will update post with SRA number when processing is complete on the NCBI end.

SRA: SRS2339870
Study: SRR5809355
BioProject: PRJNA316624
BioSample: SAMN07326085

Epinext Adaptor 1 Counts – LSU C.virginica Oil Spill Samples

Before contacting the Univ. of Oregon facility for help with this sequence demultiplexing dilemma, I contacted Epigentek to find out what the other adaptor sequence that is used in the EpiNext Post-Bisulfite DNA Library Preparation Kit (Illumina). I used grep and fastx_barcode_splitter to determine how many reads (if any) contained this adaptor sequence. All analysis was performed in the embedded Jupyter (IPython) notebook embedded below.

NBviewer: 20150317_LSU_OilSpill_EpinextAdaptor1_ID.ipynb



This adaptor sequence is not present in any of the reads in the FASTQ file analyzed.

TruSeq Adaptor Counts – LSU C.virginica Oil Spill Sequences

Initial analysis, comparing barcode identification methods, revealed the following info about demultiplexing on untrimmed sequences:

Using grep:

long barcodes: Found in ~12% of all reads

short barcodes: Found in ~25% of all reads

Using fastx_barcode_splitter:

long barcodes, beginning of line: Found in ~15% of all reads

long barcodes, end of line: Found in < 0.008% of all reads (yes, that is actually percentage)

short barcodes, beginning of line: Found in ~1.3% of all reads

short barcodes, end of line: Found in ~2.7% of all reads


Decided to determine what percentage of the sequences in this FASTQ file have just the beginning of the adaptor sequence (up to the 6bp barcode/index):


This was done to see if the numbers increased without the barcode index (i.e. see if majority of sequences are being generated from “empty” adaptors lacking barcodes).

The analysis was performed in a Jupyter (IPython) notebook and the notebook is linked, and embedded, below.

NBViewer: 20150316_LSU_OilSpill_Adapter_ID.ipynb



Using grep:

15% of the sequences match

That’s about 3% more than when the adaptor and barcode are searched as one sequence.

Using fastx_barcode_splitter:

beginning of line – 17% match

end of line – 0.06% match

The beginning of line matches are ~2% higher than when the adaptor and barcode are searched as one sequence.

Will contact Univ. of Oregon to see if they can shed any light and/or help with the demultiplexing dilemma we have here. Lots of sequence, but how did it get generated if adaptors aren’t present on all of the reads?

TruSeq Adaptor Identification Method Comparison – LSU C.virginica Oil Spill Sequences

We recently received Illumina HiSeq2500 data back from this project. Initially looking at the data, something seems off.  Using FASTQC, the quality drops of drastically towards the last 20 bases of the reads. We also see a high degree of Illumina TruSeq adaptor/index sequences present in our data.

Since this sequencing run was multiplexed (i.e. multiple libraries were pooled and run together on the HiSeq), we need to demultiplex our sequences before performing any trimming. Otherwise, the trimming could remove the index (barcodes) sequences from the data and prevent us from separating out the different libraries from each other.

However, it turns out, demultiplexing is not a simple, straightforward task. There are a variety of programs available and they all have different options. I decided to compare TruSeq index identification using two programs:

-grep (grep is a built-in command line (bash) program that searches through files to find matches to user-provided information.)
-fastx_barcode_splitter.pl (fastx_barcode_splitter.pl is a component of the fastx_tookit that searches through FASTQ files to identify matches to user-provided index/barcode sequences.)

The advantage(s) of using grep is that it’s extremely fast, easy to use, and already exists on most Unix-based computers (Linux, OS X), thus not requiring any software installation. The disadvantage(s) of using grep for a situation like this is that it is not amenable to allowing for mismatches and/or partial matches to the user-provided information.

The advantage(s) of using fastx_barcode_splitter.pl is that it can accept a user-defined number of mismatches and/or partial matches to the user-defined index/barcode sequences. The disadvantage(s) of using fastx_barcode_splitter.pl is that it requires the user to specify the expected location of the index/barcode sequence in the target sequence: either the beginning of the line or the end of the line. It will not search beyond the length(s) of the provided index/barcode sequences. That means if you index/barcode exists in the middle of your sequences, this program will not find it. Additionally, since this program doesn’t exist natively on Unix-based machines, it must be downloaded and installed by the user.

So, I tested both of these programs to see how they compared at matching both long (the TruSeq adaptor/index sequences identified with FASTQC) and “short” (the actual 6bp index sequence) barcodes.

To simplify testing, only a single sequence file was used from the data set.

All analysis was done in a Jupyter (IPython) notebook.

FASTQC HTML file for easier viewing of FASTQC output.

NBViewer version of embedded notebook below.




long barcodes: Found in ~12% of all reads

short barcodes: Found in ~25% of all reads



long barcodes, beginning of line: Found in ~15% of all reads

long barcodes, end of line: Found in < 0.008% of all reads (yes, that is actually percentage)


short barcodes, beginning of line: Found in ~1.3% of all reads

short barcodes, end of line: Found in ~2.7% of all reads


Overall, the comparison is interesting, however, the important take home from this is that in the best-case scenario (grep, short barcodes), we’re only able to identify 25% of the reads in our sequences!

It should also be noted that my analysis only used sequences in one orientation. It would be a good idea to also do this analysis by searching with the reverse and reverse complements of these sequences.