[GenoMax]

Have you tried adding "VALIDATION_STRINGENCY=LENIENT" to your picard command? Default value is STRICT and sometime it leads to picard throwing errors like this.

It does looks like reformat.sh took out some of the reads so the problem reads have been hopefully addressed.

[bwubb]

Have you tried adding "VALIDATION_STRINGENCY=LENIENT" to your picard command? Default value is STRICT and sometime it leads to picard throwing errors like this.
...

Yes indeed, but it will still be an ERROR.

However, after more tinkering I was able to come up with a combination that worked.

reformat.sh -Xmx5g in=qSortA.bam out=qSortA_reads.fq
reformat.sh -Xmx5g int=t addcolon=t uniquenames=t in=qSortA_reads.fq out=out_reads.fq
repair.sh -Xmx5g in=out_reads.fq out1=reads1.fq out2=reads2.fq

Taking those reads into alignment and so forth will pass ValidateSamFile. Im guessing these actions are allowing me to completely rewrite the pairing status, which I am perfectly ok with.

Thank you for the comments and help!

[noobie]

I want to subsample 10 million reads from both an interleaved reads file and a singletons reads file at the same time. Can reformat.sh do this?

[GenoMax]

@Noobie: I don't think so. You will have to do them independently. What is the use case here, if I may ask?

[bartn]

Hi!

I want to create a (small) bacterial RNAseq testset from transcripts using randomreads.sh:

randomreads.sh reads=30000 ref=transcripts.fasta metagenome=t out1=test_RNA_1.fq out2=test_RNA_2.fq paired=t

Mapping it back to the transcripts:

bbmap.sh ref=transcripts.fasta  in1=test_RNA_1.fq in2=test_RNA_2.fq covstats=RNA.cov

Results

Reads:                               	60000
Mapped reads:                        	54225
Mapped bases:                        	8063372
Ref scaffolds:                       	562
Ref bases:                           	521537

Percent mapped:                      	90.375
Percent proper pairs:                	64.822
Average coverage:                    	15.461
Standard deviation:                    	51.768
Percent scaffolds with any coverage: 	59.96
Percent of reference bases covered:  	45.87

Mapping it to the genome:

Reads:                               	60000
Mapped reads:                        	54104
Mapped bases:                        	8061481
Ref scaffolds:                       	1
Ref bases:                           	592529

Percent mapped:                      	90.173
Percent proper pairs:                	64.573
Average coverage:                    	13.605
Standard deviation:                    	48.628
Percent scaffolds with any coverage: 	100.00
Percent of reference bases covered:  	41.02

Why are the mapping percentages so low, what I missing with either generating the reads or mapping them back?

(bowtie2 gives an even worse result..
15482 (51.61%) aligned concordantly 0 times
14351 (47.84%) aligned concordantly exactly 1 time
167 (0.56%) aligned concordantly >1 times
)

[bartn]

randomreads.sh

Hi!

I want to created a (small) RNAseq test set from a small bacterial plasmid. So I used :

randomreads.sh reads=30000 paired=t metagenome=t  ref=transcripts.fasta out1=test_RNA_1.fq out2=test_RNA_2.fq

(Then I get 30k reads for each pair. No problem, I guess that's because the default is single end)

Then I mapped back the reads with BBmap

bbmap.sh ref=Rhizobium_Leguminosarum_plasmid_NZ_CP025014.1_transcripts.fasta  in1=Unlock_test_RNA_1.fq in2=Unlock_test_RNA_2.fq covstats=RNA.cov

Results:

Reads:                               	60000
Mapped reads:                        	54225
Mapped bases:                        	8063372
Ref scaffolds:                       	562
Ref bases:                           	521537

Percent mapped:                      	90.375
Percent proper pairs:                	64.822
Average coverage:                    	15.461
Standard deviation:                    	51.768
Percent scaffolds with any coverage: 	59.96
Percent of reference bases covered:  	45.87

The mapping percentages seem very low to me. Mapping to the genome gives similar results.
I am using randomreads.sh in the wrong way for this?

[GenoMax]

Since you asked for 30K reads you got 30K paired-end reads.

I would sugegst trying without the metagenome=t option for a single small input fasta. If you do have the entire genome then use that in your randomread generation step along with the plasmid.

[bartn]

Since you asked for 30K reads you got 30K paired-end reads.

I would sugegst trying without the metagenome=t option for a single small input fasta. If you do have the entire genome then use that in your randomread generation step along with the plasmid.

Yes, that does seem to increase the mapping percentages. Any idea why it is not working with meta option? Changes read length and insert size does not seem to do much..
The plasmid is about 0.5MB with 562 transcripts. Median transcript size is 830bp, so that could be a bit short, (but then I would expect the same with the meta option off)

I will also look for another suitable (small) genome. I don't want to add the complete genome because it should finish quick, since it's for testing purposes of tools and pipelines.

(btw, one other small thing. Adding any statistic file in the arguments will give the small summary in the stdout (like I posted before). Otherwise it will not be printed. Not sure if that is expected behavior)

[GenoMax]

Metagenome option is supposed to take in multiple (genome) sequences to generate an artificial metagenomic dataset.

BBMap writes stats to STDERR by default. So you can capture that.

[bartn]

From the description of the option I understood it can be used for RNA as well:

metagenome=f Assign scaffolds a random exponential coverage level, to simulate a metagenomic or RNA coverage distribution

Which makes sense in my view since I just give multiple transcripts instead of genomes (albeit short compared to a genome)

So why are so many reads not mapping back even thought almost all of it with an even distributed coverage does map.


What I meant with the summary part, if I run just bbmap in its simplest form:

bbmap.sh ref=transcripts.fasta  in1=read_1.fq  in2=read_2.fq

the stderr does not contain this part:

Reads:                               	120000
Mapped reads:                        	111248
Mapped bases:                        	8357095
Ref scaffolds:                       	1
Ref bases:                           	592529

Percent mapped:                      	92.707
Percent proper pairs:                	56.370
Average coverage:                    	14.104
Standard deviation:                    	46.411
Percent scaffolds with any coverage: 	100.00
Percent of reference bases covered:  	41.67

But it does with adding a covstats=cov.stats for example.

[GenoMax]

Unfortunately Brian would be the only person who can provide an authoritative answer and he no longer participates on this forum. You may want to try emailing him with your question directly.

BBMap provides a wide range of coverage/histogram/stats outputs. Default output (while voluminous does not contain coverage stats) but mainly contains % alignments which most users want to see by default. In-line help (run bbmap.sh without any options/inputs) details different options available for coverage/histogram/stats outputs.

[bartn]

That's indeed unfortunate. I might try sending an email. A previous bug report was solved very quickly!

One of the reasons I like about BBMap is indeed the wide range of statistics it can generate. Just noticed that final summary percentages like I showed, are to me the most useful thing to look at on first check was missing by default.

And thanks a lot for your answers and your time GenoMax!

[Illusive Man]

Dear BBMap users or Brian,

I have a database of about 800 proteins and all I would like to do is create an abundance table of "hits" to these 800 proteins for each one of my 30 samples. Samples as columns, genes as rows, cells as counts. Can BBMap help me create this abundance table?

Thanks

[GenoMax]

BBMap is not designed to work with protein data. You will have to find a different tool.

Take a look at blat from UCSC's Jim Kent. That should produce the stats in parse-able tab-delimited columner format. You will need to do post-processing to get the table you are looking for.

[bartn]

another program that is often used for these cases is DIAMOND

https://github.com/bbuchfink/diamond

But like GenoMax said you have to do some post processing of the output to get to the format you want..