I just came across your Lofreq paper and I'm very interested in trying this out on MiSeq data we generate here. When testing this out using Lofreq_star_2.00-rc-1 I've noticed 2 things.
In the output vcf generated I did not notice any indels. I'm curious if LoFreq2 calls indels. In the paper and earlier versions it does not, but on your blog you mention that LoFreq2 now deals with indels. Could you please verify the version I'm using deals with indels appropriately.
When running lofreq call with the --no-default-filter option the resulting vcf is still being filter by --snvqual-thresh. Is this a bug?
we've implemented indel calling in LoFreq, but are a bit reluctant to
open the feature to the public before we've extensively benchmarked it
for the paper. We will make it part of the package very soon though. I
can notify you by email if you'd like.
RE filtering: no need to worry, this is as it's supposed to be.
Normally, when you run 'lofreq call' it will execute the filter
subcommand at the end, which has default settings to remove SNVs with
high strand-bias and or are low coverage regions. You will want to get
rid of those guys, i.e. I would not recommend to use
--no-default-filter (for 'lofreq call' and likewise --no-defaults for
'lofreq filter'). If you want full control over the SNV quality see
further below for how to do this. First let me explain, why you still
see the filter command running in your verbose output: when you use
--no-default-filter as option for the call subcommand, the filter
command will be run nevertheless (with the defaults switched off
though, as you can see from the verbose output) because it has to
apply an automatically computed Bonferroni factor (applied as SNV
quality threshold). This in turn can be controlled with the -b option
of the call subcommand. For example, if you wanted to make sure to get
all SNVs (except the ones with strand-bias and at low coverage) with a
Phred quality of at least 60, you can do two things:
1. 'lofreq call -b 1 ...' followed by an explicit 'lofreq filter -B 60 ...'
or
2. 'lofreq call -b 1 -s 0.000001 ...' without explicit filtering
This latter will use a Bonferroni factor of 1 with a p-value threshold
of 1e-6 (== Phred quality 60)
I hope that answers your question. Please let me know if you need
further clarification!
I just came across your Lofreq paper and I'm very interested in trying this
out on MiSeq data we generate here. When testing this out using
Lofreq_star_2.00-rc-1 I've noticed 2 things.
In the output vcf generated I did not notice any indels. I'm curious if
LoFreq2 calls indels. In the paper and earlier versions it does not, but on
your blog you mention that LoFreq2 now deals with indels. Could you please
verify the version I'm using deals with indels appropriately.
When running lofreq call with the --no-default-filter option the resulting
vcf is still being filter by --snvqual-thresh. Is this a bug?
to elaborate on the indel calling a bit: indel calling with LoFreq is
currently a multi-stage process which is not necessarily user friendly
and requires you to know a bit about your data. Creating a simple
work-flow with all functions implemented in one binary will take
another couple of weeks.
If however your BAM file was recalibrated with GATK 2 i.e. contains
indel qualities (tags BI and BD in your BAM file) and was also indel
realigned (with e.g. GATK) then the process is rather simple: you
would need to run two binaries on your data, which will give you a
vcf-file with indel calls only. I'm happy to share those binaries with
you if you're interested. Just keep in mind this is "beta".
Let me know if you're interested via PM: wilma@gis.a-star.edu.sg
we've implemented indel calling in LoFreq, but are a bit reluctant to
open the feature to the public before we've extensively benchmarked it
for the paper. We will make it part of the package very soon though. I
can notify you by email if you'd like.
RE filtering: no need to worry, this is as it's supposed to be.
Normally, when you run 'lofreq call' it will execute the filter
subcommand at the end, which has default settings to remove SNVs with
high strand-bias and or are low coverage regions. You will want to get
rid of those guys, i.e. I would not recommend to use
--no-default-filter (for 'lofreq call' and likewise --no-defaults for
'lofreq filter'). If you want full control over the SNV quality see
further below for how to do this. First let me explain, why you still
see the filter command running in your verbose output: when you use
--no-default-filter as option for the call subcommand, the filter
command will be run nevertheless (with the defaults switched off
though, as you can see from the verbose output) because it has to
apply an automatically computed Bonferroni factor (applied as SNV
quality threshold). This in turn can be controlled with the -b option
of the call subcommand. For example, if you wanted to make sure to get
all SNVs (except the ones with strand-bias and at low coverage) with a
Phred quality of at least 60, you can do two things:
1. 'lofreq call -b 1 ...' followed by an explicit 'lofreq filter -B 60 ...'
or
2. 'lofreq call -b 1 -s 0.000001 ...' without explicit filtering
This latter will use a Bonferroni factor of 1 with a p-value threshold
of 1e-6 (== Phred quality 60)
I hope that answers your question. Please let me know if you need
further clarification!
I just came across your Lofreq paper and I'm very interested in trying this
out on MiSeq data we generate here. When testing this out using
Lofreq_star_2.00-rc-1 I've noticed 2 things.
In the output vcf generated I did not notice any indels. I'm curious if
LoFreq2 calls indels. In the paper and earlier versions it does not, but on
your blog you mention that LoFreq2 now deals with indels. Could you please
verify the version I'm using deals with indels appropriately.
When running lofreq call with the --no-default-filter option the resulting
vcf is still being filter by --snvqual-thresh. Is this a bug?
Hi there,
I just came across your Lofreq paper and I'm very interested in trying this out on MiSeq data we generate here. When testing this out using Lofreq_star_2.00-rc-1 I've noticed 2 things.
In the output vcf generated I did not notice any indels. I'm curious if LoFreq2 calls indels. In the paper and earlier versions it does not, but on your blog you mention that LoFreq2 now deals with indels. Could you please verify the version I'm using deals with indels appropriately.
When running lofreq call with the --no-default-filter option the resulting vcf is still being filter by --snvqual-thresh. Is this a bug?
Command:
lofreq call -f hg1k_v37.fasta -l intervals.bed -q 15 -Q 20 -m 15 -C 10 --no-default-filter --out lofreq.target.vcf sorted.leftAligned.bam
<snip>
Executing lofreq filter --only-passed -i /tmp/lofreq2-call-dyn-bonf.2Ut3e3 -o lofreq.target.vcf --no-defaults --snvqual-thresh 74
Thanks!
Cheers,
Rob Eveleigh
Hi Robert,
we've implemented indel calling in LoFreq, but are a bit reluctant to
open the feature to the public before we've extensively benchmarked it
for the paper. We will make it part of the package very soon though. I
can notify you by email if you'd like.
RE filtering: no need to worry, this is as it's supposed to be.
Normally, when you run 'lofreq call' it will execute the filter
subcommand at the end, which has default settings to remove SNVs with
high strand-bias and or are low coverage regions. You will want to get
rid of those guys, i.e. I would not recommend to use
--no-default-filter (for 'lofreq call' and likewise --no-defaults for
'lofreq filter'). If you want full control over the SNV quality see
further below for how to do this. First let me explain, why you still
see the filter command running in your verbose output: when you use
--no-default-filter as option for the call subcommand, the filter
command will be run nevertheless (with the defaults switched off
though, as you can see from the verbose output) because it has to
apply an automatically computed Bonferroni factor (applied as SNV
quality threshold). This in turn can be controlled with the -b option
of the call subcommand. For example, if you wanted to make sure to get
all SNVs (except the ones with strand-bias and at low coverage) with a
Phred quality of at least 60, you can do two things:
1. 'lofreq call -b 1 ...' followed by an explicit 'lofreq filter -B 60 ...'
or
2. 'lofreq call -b 1 -s 0.000001 ...' without explicit filtering
This latter will use a Bonferroni factor of 1 with a p-value threshold
of 1e-6 (== Phred quality 60)
I hope that answers your question. Please let me know if you need
further clarification!
Best,
Andreas
On 7 June 2014 03:06, Robert Eveleigh insilicool@users.sf.net wrote:
--
Andreas Wilm
andreas.wilm@gmail.com | mail@andreas-wilm.com | 0x7C68FBCC
Hi Robert,
to elaborate on the indel calling a bit: indel calling with LoFreq is
currently a multi-stage process which is not necessarily user friendly
and requires you to know a bit about your data. Creating a simple
work-flow with all functions implemented in one binary will take
another couple of weeks.
If however your BAM file was recalibrated with GATK 2 i.e. contains
indel qualities (tags BI and BD in your BAM file) and was also indel
realigned (with e.g. GATK) then the process is rather simple: you
would need to run two binaries on your data, which will give you a
vcf-file with indel calls only. I'm happy to share those binaries with
you if you're interested. Just keep in mind this is "beta".
Let me know if you're interested via PM: wilma@gis.a-star.edu.sg
Andreas
On 9 June 2014 10:57, Andreas Wilm andreas.wilm@gmail.com wrote:
--
Andreas Wilm
andreas.wilm@gmail.com | mail@andreas-wilm.com | 0x7C68FBCC