Seqanswers Leaderboard Ad



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

  • Yes .. BBMap can do that!

    BBMap suite has become an essential part of the bioinformatics tool arsenal for many ...

    Thank you Brian!

    Over the past year Brian has doled out specific advice on how to use a component from BBMap suite in standard (and not-so-standard) ways in various threads on SeqAnswers. I thought it would be useful to try and collect some "special" use cases in one thread so they could be found in one location.

    Feel free to add to this thread things I have missed. I have tried to categorize the list based on the BBMap programs. That classification is imperfect.

    There are common options for most BBMap suite programs and depending on the file extension the input/output format is automatically chosen/set.

    Note: For most programs in BBTools, you can add the parameter "config=foo.txt". Every line in "foo.txt" will be added as if it were a command-line parameter, regardless of whitespace, or the length of the file. This may be convenient if you use a certain set of parameters for different set of analyses.


    • Mapping Nanopore reads has a length cap of 6kbp. Reads longer than this will be broken into 6kbp pieces and mapped independently.

    $ -Xmx20g k=7 in=reads.fastq ref=reference.fa maxlen=1000 minlen=200 idtag ow int=f qin=33 out=mapped1.sam minratio=0.15 ignorequality slow ordered maxindel1=40 maxindel2=400

    The "maxlen" flag shreds them to a max length of 1000; you can set that up to 6000. But I found 1000 gave a higher mapping rate.

    • Using Paired-end and single-end reads at the same time

    BBMap itself can only run single-ended or paired-ended in a single run, but it has a wrapper that can accomplish it, like this:

    $ in1=read1.fq,singletons.fq in2=read2.fq,null out=mapped.sam append
    This will write all the reads to the same output file but only print the headers once. I have not tried that for bam output, only sam output

    Note about alignment stats: For paired reads, you can find the total percent mapped by adding the read 1 percent (where it says "mapped: N%") and read 2 percent, then dividing by 2. The different columns tell you the count/percent of each event. Considering the cigar strings from alignment, "Match Rate" is the number of symbols indicating a reference match (=) and error rate is the number indicating substitution, insertion, or deletion (X, I, D).

    • Exact matches when mapping small reads (e.g. miRNA)

    When mapping small RNA's with BBMap use the following flags to report only perfect matches.

    ambig=all vslow perfectmode maxsites=1000
    It should be very fast in that mode (despite the vslow flag). Vslow mainly removes masking of low-complexity repetitive kmers, which is not usually a problem but can be with extremely short sequences like microRNAs.
    • Important note about BBMap alignments

    BBMap is always nondeterministic when run in paired-end mode with multiple threads, because the insert-size average is calculated on a per-thread basis, which affects mapping; and which reads are assigned to which thread is nondeterministic. The only way to avoid that would be to restrict it to a single thread (threads=1), or map the reads as single-ended and then fix pairing afterward:

    Code: in=reads.fq outu=unmapped.fq int=f in=unmapped.fq out=paired.fq fint outs=singletons.fq
    In this case you'd want to only keep the paired output.

    BBSplit is based on BBMap, so it is also nondeterministic in paired mode with multiple threads. BBDuk and Seal (which can be used similarly to BBSplit) are always deterministic.

    • Count k-mers/find unknown primers

    $ in=reads.fq out=trimmed.fq ftr=19
    This will trim all but the first 20 bases (all bases after position 19, zero-based).

    $ in=trimmed.fq out=counts.txt fastadump=f mincount=10 k=20 rcomp=f
    This will generate a file containing the counts of all 20-mers that occurred at least 10 times, in a 2-column format that is easy to sort in Excel.

    ...etc. If the primers are 20bp long, they should be pretty obvious.
    • Convert SAM format from 1.4 to 1.3 (required for many programs)

    $ in=reads.sam out=out.sam sam=1.3
    • Removing N basecalls

    You can use BBDuk or Reformat with "qtrim=rl trimq=1". That will only trim trailing and leading bases with Q-score below 1, which means Q0, which means N (in either fasta or fastq format). The BBMap package automatically changes q-scores of Ns that are above 0 to 0 and called bases with q-scores below 2 to 2, since occasionally some Illumina software versions produces odd things like a handful of Q0 called bases or Ns with Q>0, neither of which make any sense in the Phred scale.
    • Sampling reads

    $ in=reads.fq out=sampled.fq sample=3000
    To sample 10% of the reads: in1=reads1.fq in2=reads2.fq out1=sampled1.fq out2=sampled2.fq samplerate=0.1
    or more concisely: in=reads#.fq out=sampled#.fq samplerate=0.1
    and for exact sampling: in=reads#.fq out=sampled#.fq samplereadstarget=100k
    • Changing fasta headers

    Remove anything after the first space in fasta header.

    Code: in=sequences.fasta out=renamed.fasta trd
    "trd" stands for "trim read description" and will truncate everything after the first whitespace.
    • Extract reads from a sam file

    $ in=reads.sam out=reads.fastq
    • Verify pairing and optionally de-interleave the reads

    $ in=reads.fastq verifypairing
    • Verify pairing if the reads are in separate files

    $ in1=r1.fq in2=r2.fq vpair
    If that completes successfully and says the reads were correctly paired, then you can simply de-interleave reads into two files like this:

    $ in=reads.fastq out1=r1.fastq out2=r2.fastq
    • Base quality histograms

    $ in=reads.fq qchist=qchist.txt
    That stands for "quality count histogram".
    • Filter SAM/BAM file by read length

    $ in=x.sam out=y.sam minlength=50 maxlength=200
    • Filter SAM/BAM file to detect/filter spliced reads

    $ in=mapped.bam out=filtered.bam maxdellen=50
    You can set "maxdellen" to whatever length deletion event you consider the minimum to signify splicing, which depends on the organism.
    • "Re-pair" out-of-order reads from paired-end data files

    $ in1=r1.fq.gz in2=r2.fq.gz out1=fixed1.fq.gz out2=fixed2.fq.gz outsingle=singletons.fq.gz

    BBMerge now has a new flag - "outa" or "outadapter". This allows you to automatically detect the adapter sequence of reads with short insert sizes, in case you don't know what adapters were used. It works like this:

    $ in=reads.fq outa=adapters.fa reads=1m
    Of course, it will only work for paired reads! The output fasta file will look like this:

    If you have multiplexed things with different barcodes in the adapters, the part with the barcode will show up as Ns, like this:


    Note: For BBMerge with micro-RNA, you need to add the flag mininsert=17. The default is 35, which is too long for micro-RNA libraries.
    • Identifying adapters

    If you have paired reads, and enough of the reads have inserts shorter than read length, you can identify adapter sequences with BBMerge, like this (they will be printed to adapters.fa):

    $ in1=r1.fq in2=r2.fq outa=adapters.fa


    Note: BBDuk is strictly deterministic on a per-read basis, however it does by default reorder the reads when run multithreaded. You can add the flag "ordered" to keep output reads in the same order as input reads
    • Order of operations for

    Ref thread:

    filter by minAvgQuality
    kmer-analyze (trim, filter, or mask)
    trim by overlap

    • Finding reads with a specific sequence at the beginning of read

    $ -Xmx1g in=reads.fq outm=matched.fq outu=unmatched.fq restrictleft=25 k=25 literal=AAAAACCCCCTTTTTGGGGGAAAAA
    In this case, all reads starting with "AAAAACCCCCTTTTTGGGGGAAAAA" will end up in "matched.fq" and all other reads will end up in "unmatched.fq". Specifically, the command means "look for 25-mers in the leftmost 25 bp of the read", which will require an exact prefix match, though you can relax that if you want.

    So you could bin all the reads with your known sequence, then look at the remaining reads to see what they have in common. You can do the same thing with the tail of the read using "restrictright" instead, though you can't use both restrictions at the same time.

    $ in=reads.fq outm=matched.fq literal=NNNNNNCCCCGGGGGTTTTTAAAAA k=25 copyundefined
    With the "copyundefined" flag, a copy of each reference sequence will be made representing every valid combination of defined letter. So instead of increasing memory or time use by 6^75, it only increases them by 4^6 or 4096 which is completely reasonable, but it only allows substitutions at predefined locations. You can use the "copyundefined", "hdist", and "qhdist" flags together for a lot of flexibility - for example, hdist=2 qhdist=1 and 3 Ns in the reference would allow a hamming distance of 6 with much lower resource requirements than hdist=6. Just be sure to give BBDuk as much memory as possible.
    • Removing illumina adapters (if exact adapters not known)

    If you're not sure which adapters are used, you can add "ref=truseq.fa.gz,truseq_rna.fa.gz,nextera.fa.gz" and get them all (this will increase the amount of overtrimming, though it should still be negligible).
    • Removing illumina control sequences/phiX reads

    Code: in=trimmed.fq.gz out=filtered.fq.gz k=31 ref=artifacts,phix ordered cardinality
    • Identify certain reads that contain a specific sequence

    $ in=reads.fq out=unmatched.fq outm=matched.fq literal=ACGTACGTACGTACGTAC k=18 mm=f hdist=2
    Make sure "k" is set to the exact length of the sequence. "hdist" controls the number of substitutions allowed. "outm" gets the reads that match. By default this also looks for the reverse-complement; you can disable that with "rcomp=f".
    • Extract sequences that share kmers with your sequences with BBDuk

    $ in=a.fa ref=b.fa out=c.fa mkf=1 mm=f k=31
    This will print to C all the sequences in A that share 100% of their 31-mers with sequences in B.

    • Extract sequences that contain N's with BBDuk

    Code: in=reads.fq out=readsWithoutNs.fq outm=readsWithNs.fq maxns=0
    If you have, say, 100bp reads and only want to separate reads containing all 100 Ns, change that to "maxns=99".

    General notes for

    BBDuk can operate in one of 4 kmer-matching modes:
    Right-trimming (ktrim=r), left-trimming (ktrim=l), masking (ktrim=n), and filtering (default). But it can only do one at a time because all kmers are stored in a single table. It can still do non-kmer-based operations such as quality trimming at the same time.

    BBDuk2 can do all 4 kmer operations at once and is designed for integration into automated pipelines where you do contaminant removal and adapter-trimming in a single pass to minimize filesystem I/O. Personally, I never use BBDuk2 from the command line. Both have identical capabilities and functionality otherwise, but the syntax is different.

    • Generate random reads in various formats

    $ ref=genome.fasta out=reads.fq len=100 reads=10000
    You can specify paired reads, an insert size distribution, read lengths (or length ranges), and so forth. But because I developed it to benchmark mapping algorithms, it is specifically designed to give excellent control over mutations. You can specify the number of snps, insertions, deletions, and Ns per read, either exactly or probabilistically; the lengths of these events is individually customizable, the quality values can alternately be set to allow errors to be generated on the basis of quality; there's a PacBio error model; and all of the reads are annotated with their genomic origin, so you will know the correct answer when mapping.

    Bear in mind that 50% of the reads are going to be generated from the plus strand and 50% from the minus strand. So, either a read will match the reference perfectly, OR its reverse-complement will match perfectly.

    You can generate the same set of reads with and without SNPs by fixing the seed to a positive number, like this:

    $ maxsnps=0 adderrors=false out=perfect.fastq reads=1000 minlength=18 maxlength=55 seed=5
    $ maxsnps=2 snprate=1 adderrors=false out=2snps.fastq reads=1000 minlength=18 maxlength=55 seed=5
    [As of BBmap v. 36.59] gains the ability to simulate metagenomes.

    coverage=X will automatically set "reads" to a level that will give X average coverage (decimal point is allowed).

    metagenome will assign each scaffold a random exponential variable, which decides the probability that a read be generated from that scaffold. So, if you concatenate together 20 bacterial genomes, you can run randomreads and get a metagenomic-like distribution. It could also be used for RNA-seq when using a transcriptome reference.

    The coverage is decided on a per-reference-sequence level, so if a bacterial assembly has more than one contig, you may want to glue them together first with before concatenating them with the other references.

    • Simulate a jump library

    You can simulate a 4000bp jump library from your existing data like this.

    $ cat assembly1.fa assembly2.fa > combined.fa
    $ ref=combined.fa
    $ reads=1000000 length=100 paired interleaved mininsert=3500 maxinsert=4500 bell perfect=1 q=35 out=jump.fq.gz

    $ in=ref.fasta out=reads.fastq length=200
    The difference is that RandomReads will make reads in a random order from random locations, ensuring flat coverage on average, but it won't ensure 100% coverage unless you generate many fold depth. Shred, on the other hand, gives you exactly 1x depth and exactly 100% coverage (and is not capable of modelling errors). So, the use-cases are different.
    • Demultiplex fastq files when the tag is present in the fastq read header (illumina)

    $ in=r#.fq out=out_%_#.fq prefixmode=f names=GGACTCCT+GCGATCTA,TAAGGCGA+TCTACTCT,...
    "Names" can also be a text file with one barcode per line (in exactly the format found in the read header). You do have to include all of the expected barcodes, though.

    In the output filename, the "%" symbol gets replaced by the barcode; in both the input and output names, the "#" symbol gets replaced by 1 or 2 for read 1 or read 2. It's optional, though; you can leave it out for interleaved input/output, or specify in1=/in2=/out1=/out2= if you want custom naming.

    • Plotting the length distribution of reads

    $ in=file out=histogram.txt bin=10 max=80000
    That will plot the result in bins of size 10, with everything above 80k placed in the same bin. The defaults are set for relatively short sequences so if they are many megabases long you may need to add the flag "-Xmx8g" and increase "max=" to something much higher.

    Alternatively, if these are assemblies and you're interested in continuity information (L50, N50, etc), you can run stats on each or statswrapper on all of them:

    Code: in=file

    Code: in=file,file,file,file…

    By default, "filterbyname" discards reads with names in your name list, and keeps the rest. To include them and discard the others, do this:

    $ in=003.fastq out=filter003.fq names=names003.txt include=t

    If you only know the number(s) of the fasta/fastq record(s) in a file (records start at 0) then you can use the following command to extract those reads in a new file.

    $ in=<file> id=<number,number,number...> out=<file>
    The first read (or pair) has ID 0, the second read (or pair) has ID 1, etc.

    in=<file> Specify the input file, or stdin.
    out=<file> Specify the output file, or stdout.
    id= Comma delimited list of numbers or ranges, in any order.
    For example: id=5,93,17-31,8,0,12-13
    • Splits a sam file into forward and reverse reads

    Code: mapped.sam plus.sam minus.sam unmapped.sam in=plus.sam out=plus.fq in=minus.sam out=minus.fq rcomp

    BBSplit now has the ability to output paired reads in dual files using the # symbol. For example:

    $ ref=x.fa,y.fa in1=read1.fq in2=read2.fq basename=o%_#.fq
    will produce ox_1.fq, ox_2.fq, oy_1.fq, and oy_2.fq

    You can use the # symbol for input also, like "in=read#.fq", and it will get expanded into 1 and 2.

    Added feature: One can specify a directory for the "ref=" argument. If anything in the list is a directory, it will use all fasta files in that directory. They need a fasta extension, like .fa or .fasta, but can be compressed with an additional .gz after that. Reason this is useful is to use BBSplit is to have it split input into one output file per reference file.

    NOTE: 1 By default BBSplit uses fairly strict mapping parameters; you can get the same sensitivity as BBMap by adding the flags "minid=0.76 maxindel=16k minhits=1". With those parameters it is extremely sensitive.

    NOTE: 2 BBSplit has different ambiguity settings for dealing with reads that map to multiple genomes. In any case, if the alignment score is higher to one genome than another, it will be associated with that genome only (this considers the combined scores of read pairs - pairs are always kept together). But when a read or pair has two identically-scoring mapping locations, on different genomes, the behavior is controlled by the "ambig2" flag - "ambig2=toss" will discard the read, "all" will send it to all output files, and "split" will send it to a separate file for ambiguously-mapped reads (one per genome to which it maps).

    NOTE: 3 Zero-count lines are suppressed by default, but they should be printed if you include the flag "nzo=f" (nonzeroonly=false).

    NOTE: 4 BBSplit needs multiple reference files as input; one per organism, or one for target and another for everything else. It only outputs one file per reference file., on the other hand, which is similar, can use a single concatenated file, as it (by default) will output one file per reference sequence within a concatenated set of references.
    • To generate transcript coverage stats

    $ in=mapped.sam normcov=normcoverage.txt normb=20 stats=stats.txt
    That will generate coverage per transcript, with 20 lines per transcript, each line showing the coverage for that fraction of the transcript. "stats" will contain other information like the fraction of bases in each transcript that was covered.
    • To calculate physical coverage stats (region covered by paired-end reads)

    BBMap has a "physcov" flag that allows it to report physical rather than sequenced coverage. It can be used directly in BBMap, or with pileup, if you already have a sam file. For example:

    $ in=mapped.sam covstats=coverage.txt
    • Calculating coverage of the genome

    Program will take sam or bam, sorted or unsorted.

    $ in=mapped.sam out=stats.txt hist=histogram.txt
    stats.txt will contain the average depth and percent covered of each reference sequence; the histogram will contain the exact number of bases with a each coverage level. You can also get per-base coverage or binned coverage if you want to plot the coverage. It also generates median and standard deviation, and so forth.

    It's also possible to generate coverage directly from BBMap, without an intermediate sam file, like this:

    $ in=reads.fq ref=reference.fasta nodisk covstats=stats.txt covhist=histogram.txt
    We use this a lot in situations where all you care about is coverage distributions, which is somewhat common in metagenome assemblies. It also supports most of the flags that supports, though the syntax is slightly different to prevent collisions. In each case you can see all the possible flags by running the shellscript with no arguments.
    • To bin aligned reads

    $ in=mapped.sam out=stats.txt bincov=coverage.txt binsize=1000
    That will give coverage within each bin. For read density regardless of read length, add the "startcov=t" flag.


    Dedupe ensures that there is at most one copy of any input sequence, optionally allowing contaminants (substrings) to be removed, and a variable hamming or edit distance to be specified. Usage:

    $ in=assembly1.fa,assembly2.fa out=merged.fa
    That will absorb exact duplicates and containments. You can use "hdist" and "edist" flags to allow mismatches, or get a complete list of flags by running the shellscript with no arguments.

    Dedupe will merge assemblies, but it will not produce consensus sequences or join overlapping reads; it only removes sequences that are fully contained within other sequences (allowing the specified number of mismatches or edits).

    Dedupe can remove duplicate reads from multiple files simultaneously, if they are comma-delimited (e.g. in=file1.fastq,file2.fastq,file3.fastq). And if you set the flag "uniqueonly=t" then ALL copies of duplicate reads will be removed, as opposed to the default behavior of leaving one copy of duplicate reads.

    However, it does not care which file a read came from; in other words, it can't remove only reads that are duplicates across multiple files but leave the ones that are duplicates within a file. That can still be accomplished, though, like this:

    1) Run dedupe on each sample individually, so now there are at most 1 copy of a read per sample.
    2) Run dedupe again on all of the samples together, with "uniqueonly=t". The only remaining duplicate reads will be the ones duplicated between samples, so that's all that will be removed.

    • Generate ROC curves from any aligner

    [*]index the reference

    $ ref=reference.fasta

    [*]Generate random reads

    $ reads=100000 length=100 out=synth.fastq maxq=35 midq=25 minq=15
    [*]Map to produce a sam file

    ...substitute this command with the appropriate one from your aligner of choice
    $ in=synth.fq out=mapped.sam
    [*]Generate ROC curve

    $ in=mapped.sam reads=100000
    • Calculate heterozygous rate for sequence data

    $ in=reads.fq khist=histogram.txt peaks=peaks.txt
    You can examine the histogram manually, or use the "peaks" file which tells you the number of unique kmers in each peak on the histogram. For a diploid, the first peak will be the het peak, the second will be the homozygous peak, and the rest will be repeat peaks. The peak caller is not perfect, though, so particularly with noisy data I would only rely on it for the first two peaks, and try to quantify the higher-order peaks manually if you need to (which you generally don't).

    • Compare mapped reads between two files

    To see how many mapped reads (can be mapped concordant or discordant, doesn't matter) are shared between the two alignment files and how many mapped reads are unique to one file or the other.

    $ in=file1.sam out=mapped1.sam mappedonly
    $ in=file2.sam out=mapped2.sam mappedonly
    That gets you the mapped reads only. Then:

    $ in=mapped1.sam names=mapped2.sam out=shared.sam include=t
    ...which gets you the set intersection;

    $ in=mapped1.sam names=mapped2.sam out=only1.sam include=f
    $ in=mapped2.sam names=mapped1.sam out=only2.sam include=f
    ...which get you the set subtractions.

    $ in=old.fasta out=new.fasta
    That will rename the reads as 1, 2, 3, 4, ... 222.

    You can also give a custom prefix if you want. The input has to be text format, not .doc.

    • Generating “fake” paired end reads from a single end read file

    $ in=reads.fastq out1=r1.fastq out2=r2.fastq length=100
    That will generate fake pairs from the input file, with whatever length you want (maximum of input read length). We use it in some cases for generating a fake LMP library for scaffolding from a set of contigs. Read 1 will be from the left end, and read 2 will be reverse-complemented and from the right end; both will retain the correct original qualities. And " /1" " /2" will be suffixed after the read name.

    • Generate random reads

    $ ref=genome.fasta out=reads.fq len=100 reads=10000
    "seed=-1" will use a random seed; any other value will use that specific number as the seed

    You can specify paired reads, an insert size distribution, read lengths (or length ranges), and so forth. But because I developed it to benchmark mapping algorithms, it is specifically designed to give excellent control over mutations. You can specify the number of snps, insertions, deletions, and Ns per read, either exactly or probabilistically; the lengths of these events is individually customizable, the quality values can alternately be set to allow errors to be generated on the basis of quality; there's a PacBio error model; and all of the reads are annotated with their genomic origin, so you will know the correct answer when mapping.

    • Generate saturation curves to assess sequencing depth

    $ in=reads.fq out=histogram.txt
    It works by pulling kmers from each input read, and testing whether it has been seen before, then storing it in a table.

    The bottom line, "first", tracks whether the first kmer of the read has been seen before (independent of whether it is read 1 or read 2).

    The top line, "pair", indicates whether a combined kmer from both read 1 and read 2 has been seen before. The other lines are generally safe to ignore but they track other things, like read1- or read2-specific data, and random kmers versus the first kmer.

    It plots a point every X reads (configurable, default 25000).

    In noncumulative mode (default), a point indicates "for the last X reads, this percentage had never been seen before". In this mode, once the line hits zero, sequencing more is not useful.

    In cumulative mode, a point indicates "for all reads, this percentage had never been seen before", but still only one point is plotted per X reads.


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

    In light of the quality-score issues with the NextSeq platform, and the possibility of future Illumina platforms (HiSeq 3000 and 4000) also using quantized quality scores, I developed it for recalibrating the scores to ensure accuracy and restore the full range of values.


    BBMap is designed to find the best mapping, and heuristics will cause it to ignore mappings that are valid but substantially worse. Therefore, I made a different version of it, BBMapSkimmer, which is designed to find all of the mappings above a certain threshold. The shellscript is and the usage is similar to or For primers, which I assume will be short, you may wish to use a lower than default K of, say, 10 or 11, and add the "slow" flag.

    -------------------------------------------------------------- and

    Quoted from Brian's response directly.

    I also wrote another pair of programs specifically for working with primer pairs, and will forcibly align a primer sequence (or a set of primer sequences) against a set of reference sequences to find the single best matching location per reference sequence - in other words, if you have 3 primers and 100 ref sequences, it will output a sam file with exactly 100 alignments - one per ref sequence, using the primer sequence that matched best. Of course you can also just run it with 1 primer sequence.

    So you run msa twice - once for the left primer, and once for the right primer - and generate 2 sam files. Then you feed those into, which will create a new fasta file containing the sequence between the primers, for each reference sequence. We used these programs to synthetically cut V4 out of full-length 16S sequences.

    I should say, though, that the primer sites identified are based on the normal BBMap scoring, which is not necessarily the same as where the primers would bind naturally, though with highly conserved regions there should be no difference.


    Identify type of Q-score encoding in sequence files

    $ in=seq.fq.gz
    sanger    fastq    gz    interleaved    150bp

    Newest member of BBTools. Identify constituent k-mers.
    Discussion of next-gen sequencing related bioinformatics: resources, algorithms, open source efforts, etc


    Find all k-mers for a given sequence.
    $ in=reads.fq out=kmers.txt k=4 count=t display=999
    Will produce output that looks like


    Simulate multiple mutants from a known reference (e.g. E. coli).

    $ in=e_coli.fasta out=mutant.fasta id=99 
    $ ref=mutant.fasta out=reads.fq.gz reads=5m length=150 paired adderrors
    That will create a mutant version of E.coli with 99% identity to the original, and then generate 5 million simulated read pairs from the new genome. You can repeat this multiple times; each mutant will be different.


    One can partition a large dataset with into smaller subsets (example below splits data into 8 chunks).

    Code: in=r1.fq in2=r2.fq out=r1_part%.fq out2=r2_part%.fq ways=8

    If you are concerned about file size and want the files to be as small as possible, give Clumpify a try. It can reduce filesize by around 30% losslessly by reordering the reads. I've found that this also typically accelerates subsequent analysis pipelines by a similar factor (up to 30%). Usage:

    Code: in=reads.fastq.gz out=clumped.fastq.gz
    Code: in1=reads_R1.fastq.gz in2=reads_R2.fastq.gz out1=clumped_R1.fastq.gz out2=clumped_R2.fastq.gz
    • can now mark/remove sequence duplicates (optical/PCR/otherwise) from NGS data

    This does NOT require alignments so it should prove more useful compared to Picard MarkDuplicates. Relevant options for command are listed below.

    dedupe=f optical=f (default)
    Nothing happens with regards to duplicates.
    dedupe=t optical=f
    All duplicates are detected, whether optical or not.  All copies except one are removed for each duplicate.
    dedupe=f optical=t
    Nothing happens.
    dedupe=t optical=t
    Only optical duplicates (those with an X or Y coordinate within dist) are detected.  All copies except one are removed for each duplicate.
    The allduplicates flag makes all copies of duplicates removed, rather than leaving a single copy.  But like optical, it has no effect unless dedupe=t.
    Note: If you set "dupedist" to anything greater than 0, "optical" gets enabled automatically.

    Fuse will automatically reverse-complement read 2. Pad (N) amount can be adjusted as necessary. This will for example create a full size amplicon that can be used for alignments.

    Code: in1=r1.fq in2=r2.fq pad=130 out=fused.fq fusepairs

    Following command will produce 10 output files with an equal number of sequences and no duplication.

    Code: in=X.fa out=X%.fa ways=10
    Last edited by GenoMax; 08-24-2020, 10:03 AM.

  • #2
    I feel that the best strategy to solve some very difficult problems out there in the world, like proving P = NP or creating a true artificial intelligence, would be to couch the problem in terms of read mapping and then ask Brian to add that feature to BBMap.
    Providing nextRAD genotyping and PacBio sequencing services.


    • #3
      Much thanks, GenoMax! I have added a link to this on the Sourceforge page.

      @SNPsaurus - I have no immediate plans to add artificial intelligence to BBMap, but that's a good suggestion; I'll add it to my todo list.


      • #4
        Perhaps Brian is an 'artificial intelligence' or an 'alien intelligence' who takes pity on us mere mortals to provide excellent tools for us to work with?


        • #5
          The BB map home page ( )says ...
          Short read aligner for DNA and RNA-seq data.

          Is there a special mode or feature that makes it good for RNA reads?
          Does it handle long splices especially well ?


          • #6
            Yes, that's correct. BBMap handles very long deletions or splices, making it useful for both RNA-seq or situations like fast-neutron bombardment to delete entire genes. It finds splices denovo (does not use a gene annotation file) and allows multiple introns/deletions per read. The defaults are set for plant and fungal introns, which are short; when mapping vertebrate RNA-seq data, I normally use the flag maxindel=200000 (default is 16000). It also supports the necessary flags for Cufflinks custom tags (xstag=firststrand, etc).


            • #7
              Hi Brian,
              I wouldn't say plant introns are as short as 16000. We have observed reliable introns as large as 50kbp in one species (confirmed by spliced alignments of multiple RNA-seq datasets) and 150kbp in another (maize).

              Great to hear BBmap works well on RNA-seq reads and auto detects splice sites though!


              • #8
                Oh, that's good to know. My exposure to plant RNA-seq is mainly Chlamy and Arabidopsis - and you know what they say about assumptions.


                • #9
                  Hey Brian,
                  A feature many of the bench scientists I work with would love is an automated grant proposal generator. Any chance of something like that in future updates? The input could be some images/figures and a basic subject...
                  /* Shawn Driscoll, Gene Expression Laboratory, Pfaff
                  Salk Institute for Biological Studies, La Jolla, CA, USA */


                  • #10
                    I was thinking about an automatic paper generator, which would be more useful to me. But I guess grant proposals are even more standardized. I'll see if I can roll that into Reformat.


                    • #11
                      Excellent. My PI will be stoked.
                      /* Shawn Driscoll, Gene Expression Laboratory, Pfaff
                      Salk Institute for Biological Studies, La Jolla, CA, USA */


                      • #12
                        Hi Brian,

                        Thank you for creating these awesome applications. I was wondering if you have any plans to create an OLC aligner or something that improves Dedupe so that it can merge multiple assemblies together while taking care of any duplicate sequences based on a certain or user specified percentage. Then finally it outputs a consensus sequence based on a user specified percent similarity and some other variables.

                        I believe this will be very helpful with working with viral or bacterial de-novo genome (DNA-Based) Assemblies. Some ideas.


                        • #13
                          I would also really like such a tool! Yes, I DO have plans to do something like that, but I am not sure about the timing, since I have had these plans for almost a year now. But hopefully I will make it this year, as Dedupe is really close to being an assembler, just not quite there.


                          • #14
                            quality filtering

                            Do you recommend quality trimming before mapping using BBMap? I am talking illumina 100bp PE.
                            I have been searching the forum for a while but could not find a hint.



                            • #15
                              Originally posted by fdts View Post
                              Do you recommend quality trimming before mapping using BBMap? I am talking illumina 100bp PE.
                              I have been searching the forum for a while but could not find a hint.

                              Unless you have a lot of data that has Q scores of 10 or less you can get away without quality trimming.


                              Latest Articles


                              • seqadmin
                                Exploring the Dynamics of the Tumor Microenvironment
                                by seqadmin

                                The complexity of cancer is clearly demonstrated in the diverse ecosystem of the tumor microenvironment (TME). The TME is made up of numerous cell types and its development begins with the changes that happen during oncogenesis. “Genomic mutations, copy number changes, epigenetic alterations, and alternative gene expression occur to varying degrees within the affected tumor cells,” explained Andrea O’Hara, Ph.D., Strategic Technical Specialist at Azenta. “As...
                                07-08-2024, 03:19 PM
                              • seqadmin
                                Exploring Human Diversity Through Large-Scale Omics
                                by seqadmin

                                In 2003, researchers from the Human Genome Project (HGP) announced the most comprehensive genome to date1. Although the genome wasn’t fully completed until nearly 20 years later2, numerous large-scale projects, such as the International HapMap Project and 1000 Genomes Project, continued the HGP's work, capturing extensive variation and genomic diversity within humans. Recently, newer initiatives have significantly increased in scale and expanded beyond genomics, offering a more detailed...
                                06-25-2024, 06:43 AM





                              Topics Statistics Last Post
                              Started by seqadmin, 07-19-2024, 07:20 AM
                              0 responses
                              Last Post seqadmin  
                              Started by seqadmin, 07-16-2024, 05:49 AM
                              0 responses
                              Last Post seqadmin  
                              Started by seqadmin, 07-15-2024, 06:53 AM
                              0 responses
                              Last Post seqadmin  
                              Started by seqadmin, 07-10-2024, 07:30 AM
                              0 responses
                              Last Post seqadmin