I have a tumor sample where the per base sequence quality is very low for both reads (using FATSQC). So this is what I did:
Converted the tumor.bam to fastq. (Read1.fq and Read2.fq)
bedtools bamtofq -i tumor.bam -fq Read1.fq -fq2 Read2.fqNow the FASTQC report was very poor. So I decided trimming. Here's what I used:
sickle pe -f Read1.fq -r Read2.fq -t sanger -o qtrim1.fq -p qtrim2.fq -s single.fq -q 20Now I have three files: qtrim1.fq , qtrim2.fq and single.fq. All of good quality. Now I merge them.
cat qtrim1.fq qtrim2.fq single,fq > merged.fqFinally convert to bam:
java -jar picard.jar FastqToSam merged.fq O=trimmed_bam.bam SM=tumor
The resulting bam file is very small(~460Mb) as compared to the original(~80GB). I am losing a hell lot of information doing this quality trimming. Any suggestions to get past this?
Kindly go through the steps and suggest anything wrong or something you would have done differently.
Number of reads in original bam: 1034838478
Number of reads in processed bam : 8053958
1 answer
Are you really losing a "hell of of lot information" if you are removing bad data? I would say bad data is not information, it is the opposite of information.
Perhaps the your question is:
Is my filtered data really that bad that it warrants removing it?
Answer: Usually yes, but in this case, it might be worth seeing if you are the lucky case when it is not.
Try to process it for the desired information, just the bad data alone. Align it, assemble it etc. See how it goes.
Log in to answer this question.
-
How well was the rRNA depletion for RNA-Seq experiments
written by gogeni5529 •
-
Different flagstat number after using MergeBamAlignment
written by ThomasLam •
-
How to create correct adapters for trimming (with trimmomtaic)
written by Easybella •
-
Extract reads from a bam file to have two fastq files with the same number of reads
written by macielrodriguez2 •
-
Mice driver mutations
written by Gene_MMP8
-
Remove singletons from bam
written by Gene_MMP8
-
Trimmomatic usage issue
written by Gene_MMP8
-
Tutorial: Use fastp to preprocess FASTQ data with unique molecular identifer (UMI) integrated
written by chen
-
Bulk quality filtering with bbduk
written by Felix_1993 •
-
Questions About Pair-End Reads Which Are Not Match For Bwa
written by Tonyzeng •
Was the original BAM file mapped?
FastqToSamdoesn't do anything useful, you need to remap the results. Do not merge the files like you showed, align the pairs together and the singletons by themselves.Thanks for replying. I would try your suggestion. Just one question. Converting the bam file to fastq and checking individual read quality through fastqc or directly giving the .bam file as input to fastqc and checking its quality, seems equivalent? The reason for asking this is I am getting very poor quality when I am converting to fastq while the quality seems perfectly fine when I am directly giving the bam input. Here quality means : Per base sequence quality in the fastqc report
AFAIK, fastqc reads BAM files
True. But I wanted to check the individual read qualities. Can I achieve this task by using the bam file in fastqc software?
Please use the formatting bar (especially the
codeoption) to present your post better. I've done it for you this time.Thanks. will definitely do it the next time
Hello,
why do you try to convert bam to fastq? Didn't you have the original fastq files? The read informations you get may already have been processed in some way before.
Why do you want to do quality trimming? Do you think your mapping/alignment is bad or there are to much false positives?
As Devon said FastqToSam didn't map and align your reads. This must be done again.
fin swimmer
I have just the aligned bam files from EGA. No fastq files were there. The per base sequence quality is very poor in the fastqc report.
Instead of throwing away these information you can first try to make the best out of it.
If the insertsizes are small enough, so you have a lot of overlap between the paired read, you could try the Overlap-based error-correction of bbmerge.
Also GATKs BQSR might have a good impact on your data.
fin swimmer