RNASeq: Read length different from expected

[gogodidi]

Hello all,

I have received paired-end reads for 40 samples.
The reads are supposed to be 100bp per end.
Instead, 20 of my samples are 101bp per end, while the other 20 are 100bp as expected.

Because of this, we assume that these 20 samples were all in the same lane; and that by accident there was an extra iteration in the illumina sequencing.
However, we also see a strong negative correlation with read length and quality; the samples in which we had 101bp per end lose about 30% of the reads in the trimmomatic quality step.

My question is: Has this happened to anyone else? Does it occur often, and if so, does it often affect quality? We are really quite puzzled by this.

Any ideas / clues appreciated.

Thank you!
-Thies Gehrmann

[mastal]

RNASeq: Read length different from expected

The read length of 101 bp instead of 100 bp would probably be because the operator thought there were enough reagents left to sequence the extra base.


Presumably the run that gave the 101 bp reads had some problems that resulted in poorer base qualities than the run or lane that gave the 100 bp reads.

[GenoMax]

That is odd. I do not think one can set up lanes on a single flowcell to run for different number of cycles.

Have you checked with your provider to confirm if all the samples ran together (were they multiplexed) or if there was a batch difference (2 separate runs?) that can account for your observation. I think it would probably be the latter case.

[mastal]

The read identifiers for each sample should tell you if they were on the same flow cell or lane.

[GenoMax]

The read identifiers for each sample should tell you if they were on the same flow cell or lane.

Great point. This should be easy to check.

[kmcarr]

It is somewhat standard practice for Illumina sequencing that if you want read lengths of N bases you run N+1 cycles. This has to do with the way base calling works on Illumina; to properly call the base at position n in a read you need data from cycle n+1. The last base in a read will always have a lower Q-score reflecting the added uncertainty in the base call. To mitigate this you run N+1 cycles but just report N bases per read, dropping the last, low quality base. This practice is ingrained in the Illumina run recipes; the standard PE100 recipe (with indexing) on the HiSeq runs a 209 cycles (101 + 7 + 101) adding an extra cycle each to read 1, the index read and read 2. (Interestingly the MiSeq recipes still add the extra cycle to reads 1 & 2 but not to the index read.)

Some core labs simply call and report all 101 cycles, some stick to the original practice of clipping the last base. It may be that your samples were all run together on the same flow cell for 2x101 cycles but for one set of 20 they reported all 101 cycles and the other they clipped the last base. You need to check the IDs of your samples to identify the flow cell and lanes used for each.

[gogodidi]

The read identifiers for each sample should tell you if they were on the same flow cell or lane.

Indeed, they were!
Thank you (Don't know why it didn't occur to me).

[gogodidi]

It is somewhat standard practice for Illumina sequencing that if you want read lengths of N bases you run N+1 cycles. This has to do with the way base calling works on Illumina; to properly call the base at position n in a read you need data from cycle n+1. The last base in a read will always have a lower Q-score reflecting the added uncertainty in the base call. To mitigate this you run N+1 cycles but just report N bases per read, dropping the last, low quality base. This practice is ingrained in the Illumina run recipes; the standard PE100 recipe (with indexing) on the HiSeq runs a 209 cycles (101 + 7 + 101) adding an extra cycle each to read 1, the index read and read 2. (Interestingly the MiSeq recipes still add the extra cycle to reads 1 & 2 but not to the index read.)

Some core labs simply call and report all 101 cycles, some stick to the original practice of clipping the last base. It may be that your samples were all run together on the same flow cell for 2x101 cycles but for one set of 20 they reported all 101 cycles and the other they clipped the last base. You need to check the IDs of your samples to identify the flow cell and lanes used for each.

Thank you for this information.
It is strange then, that they cut it for one lane, and not for the other.
I'll have to call them to ask about that.

Thank you so much, all of you!