Not displaying all reads for a base.

[Matthew Anderson]

Hi Jim,

I can across a situation last week with one of our users not seeing
what they expected in IGV. Which we found some behaviors of IGV some
what undesirable.

They were looking at a particular base in a bam file was only
displaying only a converge of 14 reads. Yet the coverage reported by
IGV the base was 215 and samtools count gave 215 as well. We had
disabled all filtering of reads in IGV. The region of view had a
coverage of up to 14,000.

We then tried adjusting the Alignments Maximum read depth. Which has
no affect on what IGV displayed. Is this a bug, I assumed that this is
a global setting for Alignment files.

It was not until we manually set maximum coverage depth for the file
did we see all the reads for the base. Which is fine although I would
have thought that setting the Alignments Maximum read depth would have
done the same.

The issue we had with this is that we were sorting alignments by base.
Which we expected would have displayed all the reads for the base. Not
just the ones that were loaded into view, which had us scratching our
heads why we were seeing so few reads.

So what I'm suggesting is that when sorting take into account all
reads, not just been loaded or at least the option.

Cheers Matthew

[James Robinson]

Hi Matthew,

First it looks like there is a bug here if you were only seeing 14 reads, by default the max depth is 100 and I doubt you set it that low. Could you try this in the "early access" IGV, if that is not what you were using. We will be releasing that within the next few weeks and if you repeat this with that version I would like to now.

I will investigate the global vs manual setting, the intended behavior would be for the global setting (from preferences) to take affect immediately.

The last suggestion is really not practical, if you want to see all the reads you need to increase the max depth setting. This setting and the associated downsampling is there to keep IGV from crashing from excessive depth, which it will surely do without a limit imposed.

Jim

[Matthew Anderson]

Hi Jim,

I was using 2.1 EA. It was updated to version 1954 when It loaded
today. The Global takes affect when the file loads or when the
preferences changed. Which it didn't seem to do last week.

As for the reads, I'll send you an email with some screen shots with
the behaviour I was trying to describe.


On Feb 27, 12:34 pm, James Robinson <jrobi...@broadinstitute.org>
wrote:

[Stef van Lieshout]

I've come across the same issue, IGV displaying less reads than the pop-up summary shows. Weird thing is that when I play with the "max read depth" option the number of reads jumps, sometimes showing less reads when I set a higher threshold...?

Any update / explanation for this?

ps working with version 2.1 EA (2082)


Op maandag 27 februari 2012 03:22:01 UTC+1 schreef Matthew Anderson het volgende:

[Jim Robinson]

Hi,

We've identified one situation that causes this.   Do you have large numbers of reads stacked at the same start location?     Its normal for IGV to downsample reads, but what is not normal or at least designed is to downsample beyond the "max read depth" settings.

Jim

[Jim Robinson]

Hi all,

The downsampling code in version 2.1 has been completely rewritten to be both easier to understand and control.   Rather than attempt to guarantee a particular coverage depth we now expose 2 parameters, window size and the total # of reads to load per each window.  Given a window size,  IGV will randomly downsample the reads who's start position fall in the window until the specified read count is reached, or the reads are exhausted.   Reading then proceeds to the next window and the sampling starts again.   For randomly packed reads such as a typical whole-genome sequence experiment the approximate depth for such a procedure is  == (average read length) * (reads per window) / (window length).    For other situations there is no straightforward correspondence between depth and sample size so we don't try to promise one.

There is also the option to turn sampling off altogether but we don't recommend that.  All parameters are controlled from the Preferences window.

Finally, regions where reads are downsampled are marked by a small black rectangle spanning the downsampled region just under the coverage track.  Mousing over this region will reveal the # of reads that were removed during downsampling over that region.

-- Jim