tile nomenclature

[tomc]

I am learning about Illumina (HiSeq2000?) data files by poking around a run,
and am wondering why the tiles are labelled the way they are.

My understanding is a lane is composed of 48 (non-overlapping) tiles with some number of clusters within each tile.

the tiles I see are named in a 3D 2*3*8 pattern i.e. _2308 or _1301

What is the significance of the naming convention?
How is the naming convention related to the physical order of the tiles on the lane?

Thanks Tom

[Tom Barker]

Each lane has 48 tiles as you mentioned with 3 swaths on the top surface and 3 on the bottom surface. Each swath consists of 8 tiles. I think the first number is 1 for top surface or 2 for bottom surface, the second number is the swath number 1,2 or 3 and the final number is the tile number from 1-8. For example 2307 corresponds with the bottom surface, swath 3 and the 7th tile down.

[tomc]

Thanks makes sense, I had not realised both top & bottom surfaces were imaged.

So signal from the clusters on the opposite surface of the glass are avoided
somehow, depth of field?

[kmcarr]

Thanks makes sense, I had not realised both top & bottom surfaces were imaged.

So signal from the clusters on the opposite surface of the glass are avoided
somehow, depth of field?

Yes it is done by depth of field. To expand on Tom B.'s answer Illumina typically refers to the scan swaths as 'Left', 'Middle' and 'Right'. These correspond to 1, 2 and 3 respectively as the second digit of the tile ID. The last two digits are the 'Section' within a swath (currently only 01-08 are used but this could change). The section numbers start with 1 at the output end of the flow cell and 8 is at the input end. When Illumina draws an image of a flowcell it always places the output end at the top and input end at the bottom so the section numbers run from 1 at the top to 8 at the bottom.

[tomc]

thanks for expansion.

[pmiguel]

Actually there are now 16 tiles per swath. 48 on the top surface and 48 on the bottom for a total of 96.

--
Phillip

[kmcarr]

Actually there are now 16 tiles per swath. 48 on the top surface and 48 on the bottom for a total of 96.

--
Phillip

Really, is there new software? We are running HCS 1.5.15 which I thought was the latest.

O.K. Phillip, I just saw your other post with the image from your HiSeq2000 and there are quite obviously 16 sections per swath as you said. Do you know the software version you are running? Also, the fact that Illumina had created their tile number scheme with two digits available for section number suggests they had thought about increasing the number from the start.

[kmcarr]

Went to the Illumina web site and found there is indeed an new version of HCS with explains the change from 8 to 16 tiles per swath. It is a very minor version change (1.5.15.0 to 1.5.15.1) and appears to mostly be a bug fix version. Here is the blurb from the Illumina site:

HCS 1.5.15.1/RTA 1.13/SAV 1.8 SOFTWARE
This download contains an update to the HiSeq Control Software (HCS), Real-Time Analysis (RTA) Software and Sequencing Analysis Viewing (SAV) Software for the HiSeq and HiScan SQ systems. It is recommended for any system running HCS1.5.15.0 or HCS1.4.8 and lower. For HCS1.4.8 7 users, the release updates recipes and configurations required for use when sequencing Nextera prepared libraries. For HCS1.5.15.0 users, this release fixes a memory leak causing crashes as the first Flow Cell completes. It also reduces the memory footprint of RTA to help mitigate some errors that have been occurring due to communication dropouts (the Flow Cell layout now has 16 tiles per swath instead of 8).

It sounds like the change has mostly to do with memory management issues so dividing the data into twice as many chunks means each chunk is half the size.

[pmiguel]

Yes, we were standing still for a month because of an obscure problem with our 1.4.8 install on our HiScanSQ. One of the config files needed to be updated to circumvent an edge-finding issue. This should have happened automatically, but did not happen when I updated. So we had an intermittent error that prevented us from starting runs on our HiScanSQ. It was exacerbated by the fact that this issue presented like, but was not related to, another issue well known by Tech support.

However we did not get hit with the memory leak issue. Presumably because the HiScanSQ uses the same console computer type as a HiSeq, but generates only 1/4th the amount of data. But I had the impression this was a serious issues in some shops.

Having 96 tiles/lane may induce nostalgia in people forced to working with 96 and 384 well plates to the point that they can count by 96s and 384s in their heads. (96, 192, 288, 384; 384, 768, 1152, 1536).

How many tiles per flow cell? 1536. Sadly, the number of tiles is of little significance except to HCS itself.

--
Phillip