Sequence Analysis Viewer (SAV) Trick

pmiguel · 12-08-2011, 01:05 PM

I use SAV extensively to monitor and assess a run. Especially during a run in progress. One issue I want to have a feel for is the relationship between cluster density and quality of the resultant data. Using SAV this is easy to graph. However there is an issue. For example:

Ignoring densities less than 600 K/mm2 on the graph, all looks as expected. As density increases the %>q30 drops. Above 900, things get a little crazy, with some panels dropping dramatically, while others look reasonable.

The problem is that there a set of panels that appear to be around 550 K/mm2 -- a modest density. If you look at image view these panels do not appear to have densities much lower than other panels in the lane. Instead it just looks like the cluster calling software decided to give up. Okay, irritating, but just something you deal with when pushing the upper cluster density limits.

However to get a sense of the relationship between true cluster density versus quality what is one to do? The cluster density numbers are crocked at very high densities. But how about the cluster diameters? The higher the densities, the more starved for clustering reagents a cluster will be and the more "stunted" its growth. The bizarrely named "FWHM" (Full width at half maximum) is a measure of the diameter of something roughly circular with somewhat nebulous borders. So you define the border as the average distance of a point with 1/2 the maximum intensity from the point with maximum intensity (presumably the center). How about using that?

Works great:

--
Phillip

Heisman · 12-08-2011, 01:16 PM

Not gonna lie, that's pretty cool.

csquared · 12-08-2011, 04:41 PM

FWHM is definitely a very useful metric (except in the index read). We find it most useful to compare it between all 4 bases. The more even and overlapping the plots are, the better the run.

In your run, it looks like lane 8 had a wide variation in cluster density per tile compared to lane 1 or 2 that look more uniform but still high cluster density. Your graph matches our data that 900k-950k clusters is maximizes high quality data with high output and as you go over 1M clusters, things drop off rapidly.

You probably know this but the FWHM value is not only influenced by the clusters but also the focus quality. Newer instruments with the higher megapixel cameras are better with focus/FWHM metrics/data output than older instruments.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Hadoop for NGS sequence analysis	samanta	General	0	09-03-2011 12:07 PM
illumina smallRNA adapter sequence for downstram analysis + miRNA analysis steps	ndeshpan	Bioinformatics	2	06-14-2011 10:44 PM
Sequencing Analysis Viewer trick.	pmiguel	Illumina/Solexa	1	05-04-2011 08:20 AM
Are Illumina missing a trick over cluster density?	henry.wood	Illumina/Solexa	13	01-13-2011 01:18 PM
MicroArray and Sequence Analysis	BioSlayer	Bioinformatics	4	04-06-2010 05:58 PM

12-08-2011, 01:05 PM	#1
pmiguel Senior Member Location: Purdue University, West Lafayette, Indiana Join Date: Aug 2008 Posts: 2,317	Sequence Analysis Viewer (SAV) Trick I use SAV extensively to monitor and assess a run. Especially during a run in progress. One issue I want to have a feel for is the relationship between cluster density and quality of the resultant data. Using SAV this is easy to graph. However there is an issue. For example: Ignoring densities less than 600 K/mm2 on the graph, all looks as expected. As density increases the %>q30 drops. Above 900, things get a little crazy, with some panels dropping dramatically, while others look reasonable. The problem is that there a set of panels that appear to be around 550 K/mm2 -- a modest density. If you look at image view these panels do not appear to have densities much lower than other panels in the lane. Instead it just looks like the cluster calling software decided to give up. Okay, irritating, but just something you deal with when pushing the upper cluster density limits. However to get a sense of the relationship between true cluster density versus quality what is one to do? The cluster density numbers are crocked at very high densities. But how about the cluster diameters? The higher the densities, the more starved for clustering reagents a cluster will be and the more "stunted" its growth. The bizarrely named "FWHM" (Full width at half maximum) is a measure of the diameter of something roughly circular with somewhat nebulous borders. So you define the border as the average distance of a point with 1/2 the maximum intensity from the point with maximum intensity (presumably the center). How about using that? Works great: -- Phillip

12-08-2011, 04:41 PM	#3
csquared Member Location: Huntsville, AL Join Date: May 2008 Posts: 67	FWHM is definitely a very useful metric (except in the index read). We find it most useful to compare it between all 4 bases. The more even and overlapping the plots are, the better the run. In your run, it looks like lane 8 had a wide variation in cluster density per tile compared to lane 1 or 2 that look more uniform but still high cluster density. Your graph matches our data that 900k-950k clusters is maximizes high quality data with high output and as you go over 1M clusters, things drop off rapidly. You probably know this but the FWHM value is not only influenced by the clusters but also the focus quality. Newer instruments with the higher megapixel cameras are better with focus/FWHM metrics/data output than older instruments. __________________ HudsonAlpha Institute for Biotechnology http://www.hudsonalpha.org/gsl

12-08-2011, 01:16 PM	#2
Heisman Senior Member Location: St. Louis Join Date: Dec 2010 Posts: 535	Not gonna lie, that's pretty cool.