BBMap (aligner for DNA/RNAseq) is now open-source and available for download.

[Dee Yoo]

Good day,

I am trying to use BBDuk to remove adapter and to perform quality trimming. The sequencing facility told me that they used the following:

Adapter (Index)
UDI0003

Read 1
AGATCGGAAGAGCACACGTCTGAACTCCAGTCA

Read 2
AGATCGGAAGAGCGTCGTGTAGGGAAAGAGTGT

Can you help me what command should I use to detect these adapters and to trim the reads @Q10?

Thank you!

[bpeacock44]

Hello, I am trying to use bbnorm to normalize a very large dataset - 657G per paired-end file. I've been given access to a node with 208 cores and 5888 GB RAM and I assumed it would be able to process the files, but I keep getting the following error:

Exception in thread "Thread-1407" java.lang.RuntimeException: java.io.IOException: No space left on device at stream.ReadStreamByteWriter.run(ReadStreamByteWriter.java:32)
Caused by: java.io.IOException: No space left on device (more lines follow - I can share them if they'd be helpful.)

It continues to run and later on prints again

Exception in thread "Thread-1408" java.lang.RuntimeException: java.io.IOException: No space left on device at stream.ReadStreamByteWriter.run(ReadStreamByteWriter.java:32)
Caused by: java.io.IOException: No space left on device

Then later on I get the following:

Exception in thread "Thread-1497" java.lang.RuntimeException: Writing to a terminated thread.

I ended the run at this point because I wasn't sure if the output would be viable or what exactly was happening. I was wondering a few things:

1) Should I let it keep running? Will the output be usable?
2) If the node I'm using isn't sufficient, is there a way to figure out what would be? When I ran loglog.sh, I got the following: Cardinality: 45491636111 which I calculate to mean that I'd need roughly 500 GB of RAM with the default parameters. (Assuming I made the correct calculation.)
3) Are there changes I can make to the command to lessen the load? I tried increasing the min from 5 to 10 with no success, and then I tried decreasing the hash number from 3 to 2 and then to 1 - again with no change in result.

Here is the command I used:

bbnorm.sh prefilter=TRUE in1=${WDIR}/${JB}_R1_001.clean.fastq in2=${WDIR}/${JB}_R2_001.clean.fastq out1=${WDIR}/${JB}_R1_001.norm.fastq out2=${WDIR}/${JB}_R2_001.norm.fastq target=100 min=5

I originally ran it with BBMap version 38.63 and then I tried updating it to the latest and got the same result.

Thank you so much!

[santiagorevale]

Hi Brian,

I ran filterbyname.sh to extract some FastQ reads from a bigger FastQ file and I noticed that on the quality line of the output, the # (hashtag symbol) gets replaced by an ! (exclamation mark symbol). I checked the documentation and the forums about this behaviour but couldn't find anything.

Is this expected? If so, why?

Mine were latest FastQ files produced by BaseSpace from an Illumina MiSeq instrument.

Thanks in advance.

Best,
Santiago

[Esteban Gongora]

Hi Brian Bushnell, thank you very much for developing BBMap. It is such a useful tool!

I am trying to map sequencing reads to a metagenome I had previously assembled and I am getting a very strange output. After running the alignment, I also ran pileup.sh to obtain coverage stats and I got the following output:

pileup.sh in=mapped.sam out=stats.txt

Reads: 18093134
Mapped reads: 3652730
Mapped bases: 331276393
Ref scaffolds: 134515
Ref bases: 43282856

Percent mapped: 20.188
Percent proper pairs: 12.341
Average coverage: 7.654
Average coverage with deletions: 7.689
Standard deviation: 9.676
Percent scaffolds with any coverage: 99.51
Percent of reference bases covered: 99.34

If I interpreted this summary correctly, while most of my contigs/scaffolds are being covered, only ~20% of my input reads are being mapped to said contigs. However, I do no understand what happened with the other ~80% since those were the same reads I used to assemble the contigs. Does this mean that my assembler (metaSPAdes) was not able to assemble that ~80% into any kind of contigs? And if that is the case, do you have any idea why this happens? Sorry if this last question is beyond the scope of this forum.

Thanks in advance with any help you might be able to provide,

Esteban

[KenC]

Brian Bushnell Thanks for providing this very useful toolset.

I have a few questions wrt the use of BBSplit, and I guess it's OK to put them in a single post...

[Edited: well, there has been no reply for a while and I figured out some of the answers myself, and I add them below just in case someone else run into the same problems...]

1. Is there any recommendation in terms of computation resource allocation and speeding up a run? For human+mouse genome (a total of 5.6GB .fasta files) and decompressed fastq files at about 50GB each (paired-end), I found that nearly 200GB memory was needed when using 1 thread, and it can take several days or even up to a week to run. Is this normal/expected?

[Edited: in one of my later applications where I only needed to map one of the read-pairs (decompressed at around 50G), I managed to increase the number of threads to 16, with 10G memory per thread, setting java heap size around 50G, and BBSplit finished running overnight.]

2. From the user guide I read that the `maxindel` parameter should be set to a large number for RNA-seq (e.g. 200k for human according to the guide). Another example in the guide for BBMap says:

To map vertebrate RNA-seq reads to a genome:
bbmap.sh in=reads.fq out=mapped.sam maxindel=200k ambig=random intronlen=20 xstag=us

I could not find additional information on properly different parameter settings for e.g. whole exome sequencing (WES) vs RNA-seq data. Is the above a good default for human+mouse RNA-seq while running BBSplit? (except for that intronlen is probably irrelevant for BBSplit, and xstag may be application-specific)

3. I have this problem on re-using genome index as follows:

In this post, someone asked:

After running bbsplit once using the syntax: ref=ref1.fa,ref2.fa, is it possible to re-use that index on subsequent runs using the path= parameter?

And the reply was:

Yes, it is. Also, with BBSplit, I think it will try to regenerate the index every time as long as "ref=" is specified, even if it already exists, so only do that once.

However, I have some trouble figuring out how exactly to reuse pre-built genome index when running BBSplit. Below is a toy example.

Let's say I first build an index for two reference genomes, x.fa and y.fa as follows:

bbsplit.sh build=1 ref=x.fa,y.fa path=./test

Next I would like to reuse the pre-built index for splitting many fastq files; for one pair of files, I tried, e.g.:

bbsplit.sh in1=r1.fq in2=r2.fq basename=o%_#.fq build=1 path=./test

However, when I run it, it seems that the index was re-built; contents under ./test/ref/genome/1 and ./test/ref/index/1 were removed and then re-generated from scratch. Indeed, there is this message below in the log.

NOTE: Deleting contents of ./test/ref/genome/1 because reference is specified and overwrite=true
NOTE: Deleting contents of ./test/ref/index/1 because reference is specified and overwrite=true
Writing reference.
Executing dna.FastaToChromArrays2 [ref/genome/1/merged_ref_1939486580590361.fa.gz, 1, writeinthread=false, genscaffoldinfo=true, retain, waitforwriting=false, gz=true, maxlen=536670912, writechroms=true, minscaf=1, midpad=300, startpad=8000, stoppad=8000, nodisk=false]

Set genScaffoldInfo=true
Writing chunk 1
Writing chunk 2
......

Even if I do not specify `path=./test` in both commands and just leave it to the default (i.e. the current dir), the above still happens. I also tried to specify `overwrite=f` or `ow=f`, but that did not seem to have an effect either.

Besides, I also have a related question, which is whether I can pre-build indices separately for two genomes x.fa and y.fa, saved in different paths or with different build numbers, and then later reuse them for aligning with BBSplit (potentially with something like `build=1,2` or `path=<first_path>,<second_path>`).

[Edited: I figured this out. I first build the index with `bbsplit.sh ref=x.fa,y.fa`, which generated the index in a folder named `ref`, then in the subsequent BBSplit run for each sample organized in its own sample-specific dir, I first link the `ref` folder to the sample dir, then from the sample dir run with something like `bbsplit.sh ref=x.fa,y.fa in1=r1.fq in2=r2.fq basename=o%_#.fq`, now BBSplit can find the index without any issue.]

I am using BBMap version 39.01, the latest version to date.

I appreciate your kind help in advance.

[jmw86069]

Is there anything else I need to do to test some Java changes to debug where the issue is happening?
For example, I would like to turn on verboseS=t and have it spit out debug info. I think the pairlen (MAX_PAIR_DIST) is just not being applied from what I can tell from verbose=t output.

[jmw86069]

Thank you for the response Brian Bushnell !
We are using the "pairedonly=t" flag, however the reads are still returned with PROPER_PAIR in the SAM file (sam flags 83 and 163).
It appears the MAPQ is penalized, probably due to the repetitive sequence in read2, however the reads are still shown as properly paired.

Maybe you'll spot something in the arguments below?
I've tried all combinations and all SAM alignments have the tag PROPERLY_PAIRED, and are included in the mapped output.
You can see I also added killbadpairs=t and pairedonly=t, but maybe something else has gone wrong.

bbmap.sh \
   in=testA_1.subset.fq.gz in2=testA_2.subset.fq.gz \
   ref=/reference_genomes/mm39/mm39canonical.fa \
   path=/reference_genomes/mm39/bbmap_mm39canonical/ \
   out=testA.subset.mm39.maxindel20.pairlen35000.matelen33000.mapped.bam \
   outu=testA.subset.mm39.maxindel20.pairlen35000.matelen33000.unmapped.bam \
   mappedonly=t \
   trimreaddescriptions=t \
   t=3 \
   maxindel=20 \
   strictmaxindel=t \
   inserttag=t \
   matelen=33000 \
   pairlen=35000 \
   intronlen=999999999 \
   ambig=random \
   minid=0.76 \
   killbadpairs=t \
   pairedonly=t \
   covstats=testA.subset.mm39.maxindel20.pairlen35000.matelen33000.mapped.bbmap_covstats.txt \
   covhist=testA.subset.mm39.maxindel20.pairlen35000.matelen33000.mapped.bbmap_covhist.txt \
   ihist=testA.subset.mm39.maxindel20.pairlen35000.matelen33000.mapped.bbmap_inserthist.txt \
   statsfile=testA.subset.mm39.maxindel20.pairlen35000.matelen33000.mapped.bbmap_alignmentReport.txt \
   2>&1 | tee -a output_testA.subset.mm39.maxindel20.pairlen35000.matelen33000.mapped_bbmap.txt

The SAM output is shown below, and all variations of options above appear to have no effect on the SAM flags and tags:

testA.subset.mm39.maxindel1400.pairlen1200.matelen1000.mapped.bam
QNAME                                                                FLAG  RNAME  POS       MAPQ  CIGAR                  RNEXT  PNEXT     TLEN     SEQ                                                  QUAL                                                 FLAGS
testA.subset.mm39.maxindel20.pairlen35000.matelen33000.mapped.bam
NS500489:195:HLKT3BGXX:1:12109:18917:8143                            83    chr5   11805066  2     3X4=1X2=1X4=2X2=3X29=  =      11375902  -429215  ACATGCCAAAATAACAACCCACTTTTAAGTGTATAATTCATTACACAATGC  //AE//E//E/E/E///EE///EEEEEEEEEEEEEEEEEEEEEEEEAAAAA  XT:A:R  NM:i:10  AM:i:2    X8:Z:429215
NS500489:195:HLKT3BGXX:1:12109:18917:8143                            163   chr5   11375902  3     51=                    =      11805066  429215   GACAGAGACAGAGAGAGGAACATAGACAGAGACAGAGAGAGGAACATAGAC  AAAAAEEEEEEEEEEEEEEEEEEEEEEEEEEAEEEEEEEE/E/E/E/EE/E  XT:A:R  NM:i:0   AM:i:3    X8:Z:429215
NS500489:195:HLKT3BGXX:1:11102:2182:13318                            99    chr17  20345318  42    51=                    =      20345469  202      ACTGGTGGCCTTATAACTTGTGTAATAAGAAATGTTTGAGTTTAACTATAA  AAAAAEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEE  NM:i:0  AM:i:42  X8:Z:202
NS500489:195:HLKT3BGXX:1:11102:2182:13318                            147   chr17  20345469  42    51=                    =      20345318  -202     AGCTAAGAAATAGTGAGAGCTGTGAAAATTATTCTTTCCCAGGAAGGGTAG  AEEEEAEEEEEEEEEEEEEEEEAEEEEEEEEEEE6EEEEEEE/EEEAAAAA  NM:i:0  AM:i:42  X8:Z:202

The end goal is not to include that alignment in the "mapped" output file. I can pass it through awk but that's like adding a speed bump to a racetrack, really awk-ward. Yeah, I said it.

[Brian Bushnell]

Oh, that flag might be a little misleading... it's the maximum distance for reads to be considered as 'proper pairs'. Beyond that the mapq is penalized, alternative sites are prioritized, and the 'proper pair' flag bit is not set. If you want to ban pairs from mapping at all beyond that distance you need to use either the 'killbadpairs' (which will mark one of the two as unmapped) or 'pairedonly' (which will mark them both as unmapped) flag.

[jmw86069]

Hi Brian Bushnell thanks again for the BBTools suite, it's really amazing! Lots of configurable settings, fantastic speed, logic, use of kmers, it's got it all. I'm replacing various bits of our pipelines with BBTools equivalents, and so far really loving the improvements. Also enjoyed reading through the various forum posts and responses, impressive support for the tool over the years. I digress.

I cannot figure out how to get bbmap.sh to filter paired read alignments greater than 32000 distance, which is what I understood the pairlen=32000 argument is supposed to do. It just doesn't filter out alignments. For example, one alignment shows BAM TLEN=429215 for read1 (and -429215 for read2), and there is even the optional tag X8:Z:429215. All consistent. And even with or without read lengths in the distance calculation it should be well over the cutoff (I understood pairlen was distance between read ends pointing toward each other).

Am I using the wrong argument? pairlen=32000 is default, even setting it to pairlen=40000 or pairlen=20000 it doesn't appear to do anything, or not what I expected anyway.

Thanks in advance for the help!

-James

[Brian Bushnell]

I think you found a bug ...
When you look at the given line 625 in the file ./current/shared/TrimRead.java, it evaluates the CIGAR string and asserts that it ends with a match (M or =). This should be the case because the trimReadWithMatchFast() function is only invoked for 100% match or unmapped reads:

char c=sl.cigar.charAt(sl.cigar.length()-1);
assert(c=='M' || c=='=') : c+"; "+sl.cigar+"\n"+sl;

But in your case, you have the hard-trimmed right part, so c is a H and the assertion fails, even though the function is rightfully invoked, since your remaining 22 bases are all matching.

Yes, that is exactly correct; good analysis. I need to add support for H symbols, which I did not expect to encounter in that function.

[Thias]

I think you found a bug ...
When you look at the given line 625 in the file ./current/shared/TrimRead.java, it evaluates the CIGAR string and asserts that it ends with a match (M or =). This should be the case because the trimReadWithMatchFast() function is only invoked for 100% match or unmapped reads:

char c=sl.cigar.charAt(sl.cigar.length()-1);
assert(c=='M' || c=='=') : c+"; "+sl.cigar+"\n"+sl;

But in your case, you have the hard-trimmed right part, so c is a H and the assertion fails, even though the function is rightfully invoked, since your remaining 22 bases are all matching.

[CharlesARoy]

Hi Brian Bushnell

I recently installed BBMap_38.98.tar.gz from SourceForge on a machine running Ubuntu 20.04.4 LTS.

Per the usage guide, I checked my current version of Java with "java -Xmx90m -version" which returned the following:

openjdk version "11.0.15" 2022-04-19
OpenJDK Runtime Environment (build 11.0.15+10-Ubuntu-0ubuntu0.20.04.1)
OpenJDK 64-Bit Server VM (build 11.0.15+10-Ubuntu-0ubuntu0.20.04.1, mixed mode, sharing)

I tried running reformat on a sam file with a command like:

reformat.sh in=in.sam out=out.sam forcetrimleft=10

And got the following error:

java -ea -Xms300m -cp /data/usr/croy/tools/bbmap/current/ jgi.ReformatReads in=in.sam out=out.sam forcetrimleft=10
Executing jgi.ReformatReads [in=in.sam, out=out.sam, forcetrimleft=10]

Input is being processed as unpaired
Waiting on header to be read from a sam file.
Exception in thread "main" java.lang.AssertionError: H; 22M123H
FS10001085:123:BRL95608-3321:1:1101:7680:1000 2195 chr3 50608025 18 22M123H = 50607939 -98 AGTGGTTTTCACTGACAGCGTG F,:FFFFFFFF::FFF:F,F,F NM:i:0 MD:Z:22 MC:Z:93M21D52M AS:i:22 XS:i:0 SA:Z:chr3,50608053,-,17S128M,60,1;
at shared.TrimRead.trimReadWithMatchFast(TrimRead.java:625)
at shared.TrimRead.trimReadWithMatch(TrimRead.java:684)
at shared.TrimRead.trimByAmount(TrimRead.java:279)
at shared.TrimRead.trimToPosition(TrimRead.java:240)
at jgi.ReformatReads.process(ReformatReads.java:919)
at jgi.ReformatReads.main(ReformatReads.java:53)

The sam file was aligned to hg38 using a recent version of BWA MEM (v0.7.17-r1188). The CIGAR string seems to follow the SAM specs as far as I can tell, so I'm not sure where things are going wrong.

Please let me know if you need any more information.

Best,

Charles

[Omri.huji]

Is bbmap missing mappings of bbmapskimmer ?

Hello Brian,

I will be glad to get your help with the next issue.
I mapped a large DB of reads(over 100K) against the human genome, and found different results between bbmap's and bbmapskimmer's mappings.

I used bbmap to map reads against hg19, then I was interested finding all the mappings over the genome- so I used bbmapskimmer.
Except this concept I do not know any main differences between bbmap and bbmapskimmer codes (am I right?).

And still, bbmap mapped 98,425 reads and bbmapskimmer mapped 107,708. While 96,987 reads are shared.
Means bbmap mapped 1,438 reads that wasn't mapped by bbmapskimmer, which did mapped another 10,721 that wasn't mapped by bbmap- some of them with perfect minid score of 1.0.
Is there any possible reason for that?

I attached SAM files, with the relevant mappings for reads with maximum minid score in each algorithem (best minid for 1,438 reads of bbmap is 0.96, and for bbmapskimmer is 1.0).


- I used the next reference genome: hg19_main_mask_ribo_animal_allplant_allfungus.fa.gz

- All mappings were done with next parameters: minid=0.9 ambig=all

[sghignone]

filter reads based on GC% content

Hello Brian and everyone,
I'm wondering whether it could be possible (and with which tool) to perform a read filter based on its GC% content.
For example, working with a metagenome with fungal and bacterial DNA fragments, how can I retain only paired-end reads with a GC > 50% and isolate bacterial reads?
Thanks for the attention
Stefano

[ellalalalala]

Error running BBMap

Hi!

I already used BBMap successfully in the past to map RNA-seq data to a multi-FASTA transcriptome. Now I was trying to map to the human genome (downloaded from human genome browser of NCBI). I used the following command:

bbmap.sh ref=$mapping_ref build="$BBmap_ref_ID" in=$BBduked_reads1 in2=$BBduked_reads2 \
outu=unmapped_"$name1"_"$date".sam outm=mapped_"$name1"_"$date".sam \
maxindel=200k mdtag=true sam=1.4 subfilter=3 pairedonly=true

Unfortunately I´m running into an error:

[...]
Retaining first best site only for ambiguous mappings.
Writing reference.
Executing dna.FastaToChromArrays2 [/scratch/hpc-prf-ptma2/ptma2001/bbmap/GRCh38_genome_16032022.fna, 6, writeinthread=false, genscaffoldinfo=true, retain, waitforwriting=false, gz=true, maxlen=536670912, writechroms=true, minscaf=1, midpad=300, startpad=8000, stoppad=8000, nodisk=false]

Set genScaffoldInfo=true
Writing chunk 1
Writing chunk 2
Exception in thread "main" java.lang.NumberFormatException: For input string: "ӌ�jHN�H��¬���l��,ΚK��/S��ͩm���T
�!�l��X��o�ã’*s$�s^��m�h
J@�
�m��$֧����@�H,���xI%_ Uz�?� S�B�2[*i�_����)i�NEր�ٜ�?Z�i£�� "
at java.lang.NumberFormatException.forInputString(NumberFormatException.java:65)
at java.lang.Integer.parseInt(Integer.java:580)
at java.lang.Integer.parseInt(Integer.java:615)
at dna.Data.setGenome2(Data.java:1016)
at dna.Data.setGenome(Data.java:769)
at align2.BBMap.loadIndex(BBMap.java:316)
at align2.BBMap.main(BBMap.java:32)

Any idea what causes the problem and how I could fix it?

Thanks a lot in advance and have a great day

Ella