We have about 1.2 Gb of PacBio sequences for a fish genome along with plenty of Illumina data. So far, we processed the PacBio reads using BLASR and our home-cooked kmer-based algorithms, but yesterday we decided to give PacBioToCA a try after reading encouraging e-mail comments from Lex Nederbragt. While the program is still running, here are bits and pieces of information for others interested in running it.
What does PacBioToCA stand for?
CA stands for Celera Assembler. The code is expected to convert PacBio reads to reads in CA format (‘frg’), but it also produces fasta and fastq versions of error-corrected reads. For most of our data, frg file is essentially a wrapper on fasta and fastq libraries.
How to run?
The sourceforge webpage on PacBioToCA is good enough to let you get started. If you do not have Celera assembler, you will need to install it first. You can either copy the binary files from Sourceforge, or download source code and compile Celera assembler yourself.
After installing Celera assembler, please follow these steps to clean PacBio reads:
i) PacBioToCA cleans PacBio reads based on a clean set of reads (Illumina, 454). So, you will need to create a folder and copy PacBio reads and clean reads there.
ii) If your PacBio reads are in fasta format, you will need to create fastq by using convertFastaAndQualToFastq utility. It is a simple program and adds two extra lines after every fasta sequence entry. All quality scores are marked as ‘I’.
iii) You will need to create a frg wrapper file for your clean reads. The script FastqtoCA allows you to create the frg wrapper file.
iv) You will need to create a pacbio.spec file. To start with, you can copy this file and get started. It is prepared for high-memory multi-core machines. If you are running on a grid, use a different file.
Once all files are ready, please run -
pacBioToCA -l test -t 32 -s pacbio.spec -fastq pacbio.fastq clean.frg
-l test = is the name, where you want your output
-t 32 = number of threads
-s pacbio.spec = spec file listed in (iv) above
-fastq pacbio.fastq = the PacBio file you want to get cleaned
clean.frg = frg wrapper for clean Illumina or 454
The above command will allow you to get started and you will receive your clean PacBio data in about 300 days !! (no kidding)
Where can things go wrong?
We found a highly informative SeqAnswers thread with comments from many bioinformaticians, who tried PacBioToCA script. There were many complaints regarding speed of the program.
I’m also running into speed issues. My 50x coverage was estimated to take 4weeks to complete.
I would be surprised if all the people using it are ready to wait that long
Above one is not the worst. Another user complained -
I also tried to run the pacbioToCa pipeline, but for our case it initially took 200days to complete. It turned out that we have a huge E.coli contamination, making the coverage of that genome over 5000x.
We are puzzled that someone waited for 200 days before stopping the script and trying to figure out what went wrong !!
There were other hopeful comments such as from one user, who reduced the running time from 8 days to 1 hour (!!!) by modifying spec file slightly. A modified spec file is posted in the above thread.
How does PacBioToCA work?
Above comments prompted us to take a look at the code for PacBioToCA to make sense of what is going on. We do not fully understand it yet, but here is a rough sketch.
1. PacBioToCA is a 514 line PERL wrapper script, which calls another large PERL script called runCA to do first part of the analysis. What is runCA? It stands for ‘run Celera Assembler’.
3. Then runCA creates a script ‘mertrim.sh’ that calls ‘merTrim’ executable to trim good reads based on k-mer frequency. The executable merTrim compiles from /wgs-7.0/src/AS_MER directory of Celera assembler.
4. Script runCA creates a script called ‘overlap.sh’ that calls overlapInCore next to determine overlaps. This executable compiles from C-code in ‘wgs-7.0/src/AS_OVM’ directory of Celera Assembler).
/home/PacBioToCA/Linux-amd64/bin/overlapInCore -G --hashbits 24 --hashload 0.75 -t 2 -h 13000006-13200005 -r 1-50000000 --hashstrings 200000 --hashdatalen 20000000 -k 14 -k /home/PacBioToCA/working/pacbiotoca/temptry/0-mercounts/asm.nmers.obt.fasta -o /home/PacBioToCA/working/pacbiotoca/temptry/0-overlaptrim-overlap/001/000066.ovb.WORKING.gz -H 1-1 -R 1-1 /home/PacBioToCA/working/pacbiotoca/temptry/asm.gkpStore
overlapInCore -G --hashbits 24 --hashload 0.75 -t 2 -h 13000006-13200005 -r 1-50000000 --hashstrings 200000 --hashdatalen 20000000 -k 14 -k temptry/0-mercounts/asm.nmers.obt.fasta -o temptry/0-overlaptrim-overlap/001/000066.ovb.WORKING.gz -H 1-1 -R 1-1 temptry/asm.gkpStore
For our data, ‘overlap.sh’ will be run 19798 times and only 66 of them complete so far in a day. So, we are looking at 299 days of wait time.
Edit. On second look, things are not that bad. Much of the time was spent on running Meryl for k-mer counting. The script overlap.sh started to run only today, and completed 116 in 4 hours. So, we are looking at 29 days, not 299 days.
5. Finally PacBioToCA calls correctPacBio, another executable that compiles from wgs-7.0/src/AS_PBR/CorrectPacBio.cc.
Adventurous users may separately run above components of PacBioToCA on small test data, and then write their own pipeline optimized for their computing hardware (# of cores, RAM, etc.). If we are successful, we will let you know.