Cluster position units in fastq file

Question

[rbierman]

I have a fastq file with headers that include positions in the same format as the wikipedia example for fastq:

@HWUSI-EAS100R:6:73:941:1973#0/1

I know that (x,y) is (941,1973), but what are the units? Microns? Some illumina unit?

If there was another cluster at (931, 1973) would they be 10 microns apart?

Thanks

[GenoMax]

AFAIK there is no unit for those numbers since the clusters form ad hoc on a non-patterned flowcell.

[rbierman]

Thank you for your comment. I don't see why clusters forming ad hoc on a non-patterned flowcell would prohibit measurement. Since the readout is microscopy wouldn't using the known magnification and observed distance between clusters and a defined origin allow you to convert to known units?

[GenoMax]

That is certainly possible. I have not thought about those x,y positions being assigned specific units nor have seen them being assigned specific units. You may want to email techsupport at illumina.com with this question to get an official answer. Post their response here.

[Pierre Lindenbaum]

I'm suprised there is no unit because the distance between the clusters is used to detect the optical duplicons (?).

[GenoMax]

@Brian implemented a x+n y+n metric to identify optical duplicates (n is different for each type of sequencer) in clumpify.sh. He has referred to n as pixels in his post but I am not sure if that is based on concrete information or by practical examination of past datasets with identical sequence clusters within a n distance of each other.

[rbierman]

Thanks for the link! My understanding is that @Brian would have no need to convert from pixels to physical units in clumpify.sh, but this does suggest that the positions do have a physical intepretation

[rbierman]

I also opened this question on stack overflow and have more details there (link https://stackoverflow.com/questions/49182629/cluster-position-units-in-fastq-file). TL;DR The required measurements are proprietary, but using the cluster densities provided you can make estimates of the number of pixels in a micron.

Still very interested if someone has more knowledge

[GenoMax]

Curious as to why you are interested in this? Perhaps some other work around may exist.