Is the flowcell's layout known to affect perceived cluster density on a given tile?

[der_eiskern]

I just did my first sequencing run on the GAIIx. I was doing some quality control calculations from the data contained in the Summary File and I noticed something that I didnt' expect to see.

i had very clear peaks (maxima) in the cluster density around Tile 1 and Tile 120 with a minima in the middle. In other words, the graph of Tile # (X-axis) versus Cluster Density (Y-axis) looks like a parabola.

For lanes whose cluster density ranged from 150,000 to 220,000 the amount of mappable sequence was proportional (though not linearly) to the cluster density; while for the lanes that had between 230,000 and 250,000 clusters per tile, the trend seemed to be inversely proportional.

As expected, Lane 8 was yielded <50% mappable reads of the ones that passed filter with a large amount not passing filters but it clearly had this parabolic distribution across the lane's tiles.

My question is, does the parabolic cluster density reflect problems with the optics of detecting clusters, a problem with the flowcell, or a problem with the pump on the cluster generation station? or is this a normal pattern in your experience? Secondly, what order are the tiles on the flowcell? is it a 2x60 or a 1x120 or 3x40 design for each lane?

Thanks,
der.

PS. i should clarify that these numbers are before QC. the numbers after QC are not parabolic when graphed.

[kmcarr]

Secondly, what order are the tiles on the flowcell? is it a 2x60 or a 1x120 or 3x40 design for each lane?

The tiles for each lane are arranged in 2 columns of 60 tiles. The numbering starts with tile 1 at the bottom (nearest the fluid inlet) up to 60 at the top then shifts over to column 2 and continues 61-120 from top to bottom. Tiles 1 and 120 are next to each other. Your "parabola" is telling you that the clusters are more dense at the ends of the lane nearest the inlet holes.

It is not surprising that the amount of useable data from lanes with >230,000 clusters per tile is proportionately lower. The recommended cluster density is ~200-210 thousand clusters per tile. Beyond that there is more signal overlap between tiles which lowers the quality of the base calls.

[james hadfield]

I always supposed that we see cluster denisity gradients since the DNA we are hybridising is reducing in concentration as it flows down the flow cell and DNA 'sticks' to the oligos on the flow cell. This was always the basis of our smiley face showing that the flowcell was the right way round in the instrument! A sad face inicated that someone had made a mistake.