Seqanswers Leaderboard Ad

Collapse

Announcement

Collapse
No announcement yet.
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Is trimming necessary before mapping SOLiD data with BFAST?

    Hi,

    I am wondering whether trimming is necessary if using BFAST for aligning SOLiD reads? BFAST has been pretty powerful in mapping cs reads accurately while tolerating mismatches and indels - ~90% of my reads (single-end, 75bp, from genomic DNA) were uniquely mapped with best alignment score.

    However, the base quality drops dramatically at 3' end of the SOLiD reads. Here are the base quality distributions of raw reads (csfastq format) and mapped reads (bam format)

    - Raw reads: I suspect it detects the wrong quality format - Illumina not Sanger - though (not sure)
    Code:
    fastqc $reads.csfastq


    - BFAST mapped reads: I think this time it detects the correct quality format - Sanger
    Code:
    fastqc $reads.mapped.srt.rmdup.bam


    If mapping without trimming, the low-quality 3' end of the reads are mapped, but with lots of mismatches. Here is a snapshot from the bam file in GenomeView (all mismatches are at the 3' end of the mapped reads)

    - green: Read mapped to the forward strand
    - blue: Read mapped to the reverse strand


    I am debating whether to trim the reads before mapping, in the purpose of accurate multi-sample variant calling, as my preliminary analysis (without re-align or re-cal) using varscan returned lots of false positive snps, which are located in the the noisy mapping regions....on the other hand, I would like not to throw out reads that are indeed mapped (even with mismatches).

    here are the thresholds that I set for variant calling:
    Code:
    java -Xmx4g -jar VarScan.v2.2.11.jar  mpileup2snp \
    $reads.mapped.srt.rmdup.merged.mpileup \
    --min-coverage 8 \
    --min-reads2 2 \
    --min-var-freq 0.2 \
    --min-avg-qual 20 \
    --p-value 0.01 \
    > $reads.snp
    Any help is highly appreciated! Much thanks in advance.

    Sonia
    Last edited by sonia.bao; 06-01-2012, 09:21 PM.

  • #2
    You should definitely try a re-aligner like SRMA to clean up the alignments, especially for indel-calling. You could also try a higher cutoff for base quality (e.g. -Q 20) or do some post-alignment filtering (e.g. remove reads with lots of mismatches, low mapping quality or low average quality score) to get more reliable results.

    Comment


    • #3
      I have found it to be fairly advantageous to trim reads in order to obtain a higher percentage of mapped reads since the ambiguous mapping percentage isn't significantly different, at least in my data sets.

      However, when calling variants using varscan, I have found it to be fairly pointless to trim the reads, though it is probably dependent upon the average coverage of you reference. For one example, I had 35-50 paired-end shotgun reads on a ~4MB genome with ~300x average coverage. I tried trimming and filtering and it did not produce a significant number of different SNV counts (as seen below). I didn't use BFAST, though I did use bowtie and shrimp.

      Unfiltered untrimmed: 11697
      Untrimmed filtered: 11198
      Trimmed filtered: 11340
      Trimmed unfiltered: 11366

      I've also notice a drastic difference in assembly for those that are crazy enough to go de novo with solid.

      I did notice, however, that when using varscan, turning the strand filter off eliminated a lot of the false positives. Also note that it may act different on indels as I have not tried it. However, I have the data ready so I can try it out and let you know. I hope that this helps!
      Last edited by twaddlac; 06-02-2012, 08:45 AM. Reason: Forgot to mention assembly!

      Comment


      • #4
        This thread may also be of interest to you:

        Discussion of next-gen sequencing related bioinformatics: resources, algorithms, open source efforts, etc

        Comment


        • #5
          Originally posted by Chipper View Post
          You should definitely try a re-aligner like SRMA to clean up the alignments, especially for indel-calling. You could also try a higher cutoff for base quality (e.g. -Q 20) or do some post-alignment filtering (e.g. remove reads with lots of mismatches, low mapping quality or low average quality score) to get more reliable results.
          Thank you Chipper! I tried re-align with GATK and SRMA, and the alignment was not improved much. Here is a snapshot of SRMA output (same as GATK output):



          I am not sure is that because I have a small genome (~4MB) with ~10M mapped reads only?

          Any comments and suggestions are highly appreciated!
          Last edited by sonia.bao; 06-02-2012, 07:25 PM. Reason: re-align finished

          Comment


          • #6
            Originally posted by twaddlac View Post
            I have found it to be fairly advantageous to trim reads in order to obtain a higher percentage of mapped reads since the ambiguous mapping percentage isn't significantly different, at least in my data sets.

            However, when calling variants using varscan, I have found it to be fairly pointless to trim the reads, though it is probably dependent upon the average coverage of you reference. For one example, I had 35-50 paired-end shotgun reads on a ~4MB genome with ~300x average coverage. I tried trimming and filtering and it did not produce a significant number of different SNV counts (as seen below). I didn't use BFAST, though I did use bowtie and shrimp.

            Unfiltered untrimmed: 11697
            Untrimmed filtered: 11198
            Trimmed filtered: 11340
            Trimmed unfiltered: 11366

            I've also notice a drastic difference in assembly for those that are crazy enough to go de novo with solid.

            I did notice, however, that when using varscan, turning the strand filter off eliminated a lot of the false positives. Also note that it may act different on indels as I have not tried it. However, I have the data ready so I can try it out and let you know. I hope that this helps!
            Thank you very much twaddlac! This is of great help Will try varscan again and get back here.

            I am curious about why turning the the strand filter off reduced the false positive rate. Isn't it set for filtering out extreme strand bias?

            Would you mind sharing which tools you used for trimming SOLiD reads? I saw people using cutadapt and clc bio but have not tried them yet. Thanks much!
            Last edited by sonia.bao; 06-02-2012, 06:31 PM.

            Comment


            • #7
              Originally posted by twaddlac View Post
              This thread may also be of interest to you:

              http://seqanswers.com/forums/showthread.php?t=16745
              Thank you very much! The thread is very useful. I've got similar results with various mappers too - BFAST and LifeScope map the most, and bowtie works best for the high-quality reads (the most "clean" alignment). We finally decided to go with BFAST, as LifeScope was a pain to install on the server (and runs relatively slow), and novoalignCS was too slow with 1 thread only (the free version) though it gave pretty good results.

              BFAST has been good at aligning reads from genomic DNA, haven't tried it with RNA-Seq though....
              Last edited by sonia.bao; 06-02-2012, 06:36 PM.

              Comment


              • #8
                I am curious about why turning the the strand filter off reduced the false positive rate. Isn't it set for filtering out extreme strand bias?
                Sorry, I need to mind my terminology more. It didn't produce false positives per se - it mainly provided more junk than without it. However, I found it advantageous to turn it off when using VarScan on my datasets. I'd be interested to hear how it works for you!

                Would you mind sharing which tools you used for trimming SOLiD reads?
                I use SOLiD_preprocess_meanFilter_SOPRA_v1.pl which comes along with the Sopra package (http://www.physics.rutgers.edu/~anirvans/SOPRA/). I used the defaults but there are many tuning parameters, too. This also does the trimming if you opt for it.

                Comment


                • #9
                  Hi everyone on this great thread!

                  What did you en up using to trim your SOLiD reads? I've read elsewhere that the chemistry of SOLiD gives for a very weird-looking per-base quality boxblot in FastQC -- so does anybody know what is the standard for a good quality run?

                  Thanks!!

                  Carmen

                  Comment


                  • #10
                    For the SOLiD 4 I've had good results with 35bp and 50bp with the SOLiD 5. As I recall, especially with the SOLiD 4, those are where the noticeable quality drop-offs were.

                    Comment


                    • #11
                      Thanks, twaddlac!

                      Would you mind sharing what program ou use to trim your reads?

                      Thanks,

                      Carmen

                      Comment


                      • #12
                        Originally posted by twaddlac View Post
                        I use SOLiD_preprocess_meanFilter_SOPRA_v1.pl which comes along with the Sopra package (http://www.physics.rutgers.edu/~anirvans/SOPRA/). I used the defaults but there are many tuning parameters, too. This also does the trimming if you opt for it.
                        It's a perl script that came with the SOPRA download - you don't have to install it to use it.

                        Comment


                        • #13
                          Cool.

                          For anyone out there looking for a way to do this, here is the simple command that will help you out in case you already have your .csfasta and .qual files merged into a single colorspace fastq file:

                          To keep from base 1 and length 25:

                          awk 'NR % 2 == 0 { print substr($1, 1, 25) } NR % 2 == 1' SRR388230_2.fastq > SRR388230_2_TRIM.fastq

                          It doesn't take too long to run!

                          Comment

                          Latest Articles

                          Collapse

                          • seqadmin
                            Strategies for Sequencing Challenging Samples
                            by seqadmin


                            Despite advancements in sequencing platforms and related sample preparation technologies, certain sample types continue to present significant challenges that can compromise sequencing results. Pedro Echave, Senior Manager of the Global Business Segment at Revvity, explained that the success of a sequencing experiment ultimately depends on the amount and integrity of the nucleic acid template (RNA or DNA) obtained from a sample. “The better the quality of the nucleic acid isolated...
                            03-22-2024, 06:39 AM
                          • seqadmin
                            Techniques and Challenges in Conservation Genomics
                            by seqadmin



                            The field of conservation genomics centers on applying genomics technologies in support of conservation efforts and the preservation of biodiversity. This article features interviews with two researchers who showcase their innovative work and highlight the current state and future of conservation genomics.

                            Avian Conservation
                            Matthew DeSaix, a recent doctoral graduate from Kristen Ruegg’s lab at The University of Colorado, shared that most of his research...
                            03-08-2024, 10:41 AM

                          ad_right_rmr

                          Collapse

                          News

                          Collapse

                          Topics Statistics Last Post
                          Started by seqadmin, Yesterday, 06:37 PM
                          0 responses
                          10 views
                          0 likes
                          Last Post seqadmin  
                          Started by seqadmin, Yesterday, 06:07 PM
                          0 responses
                          9 views
                          0 likes
                          Last Post seqadmin  
                          Started by seqadmin, 03-22-2024, 10:03 AM
                          0 responses
                          51 views
                          0 likes
                          Last Post seqadmin  
                          Started by seqadmin, 03-21-2024, 07:32 AM
                          0 responses
                          67 views
                          0 likes
                          Last Post seqadmin  
                          Working...
                          X