5 replies to this topic

[Clare]

Hi everyone

I hope someone can help out with this otherwise we have just wasted a %^&* load of money on arrays!!

A while ago we did a 'pilot' microRNA expression array (Illumina) consisting of 12 samples. Recently, we did 36 more. When we try to cluster the data however (after normalisation - see below), we get two distinct clusters based on when they were run

Bioinformatician notes: "I appended the second
>> file to the first one and loaded the combined file into R for processing
>> with the beadarray package. I tried both quantile normalization and vsn.
>> The between-array normalization seemed to have pretty good results in
>> terms of equalizing the expression value distribution for the samples
>> however, hierarchical clustering revealed the samples to be clustered
>> based on when they were run".

Has anyone else had this problem? We've had no problemss with our Illumina GEX, just the microRNA arrays. They are supposed to be reproducible (according to Illumina) but the head of the facility that runs ours arrays says they are not. I wish she had told us this when we did the pilot!!!  

Any advice would be most welcome!!!

Clare

[Mr. Naive]

Did you mean that the clustering was not seen as control vs treatment but showed with date clustering? If so, may be a technical screw-up. Check the lab notes that was written. Sorry, i dont know much. I started recently.

[Clare]

Yes, date clustering We now think it's a problem with the arrays and that's why Illumina discontinued them!! I am (*&^%£ off because Illumina assured us they were ok...but one of the controls on the array has NEVER worked (!"!!).

Anyway, as we had 12 more arrays left, I have submitted more samples to see if we get the date effect again (or perhaps our pilot experiment was just rubbish).

Clare

Mr. Naive, on May 27 2010, 11:41 PM, said:

Did you mean that the clustering was not seen as control vs treatment but showed with date clustering? If so, may be a technical screw-up. Check the lab notes that was written. Sorry, i dont know much. I started recently.

[onclab]

Clare, on Jun 2 2010, 12:51 PM, said:

Yes, date clustering sad.gif We now think it's a problem with the arrays and that's why Illumina discontinued them!! I am (*&^%£ off because Illumina assured us they were ok...but one of the controls on the array has NEVER worked (!"!!).

Anyway, as we had 12 more arrays left, I have submitted more samples to see if we get the date effect again (or perhaps our pilot experiment was just rubbish).

Clare

Hi,

I haven't run Illumina's MicroRNA chips but I have performed profiling using the GEX chips, where we have performed several pilot projects at different time points. If we wanted to be able to compare results from two different runs, then we would usually include a few of the samples from a previous run along with the present run...just to see if there is any technical variation. I'm not sure if you have already tried this, but it could be worth a shot.

Good luck!

Edited by onclab, 09 June 2010 - 10:16 AM.

[Clare]

We ran another 12 samples on the same platform, including samples that were run on the first 2 experiments. Samples that were run on both the 2nd and 3rd experiment cluster together nicely - so for some reason the pilot experiment we did just didn't work

Clare

onclab, on Jun 9 2010, 07:15 PM, said:

Clare, on Jun 2 2010, 12:51 PM, said:

Yes, date clustering sad.gif We now think it's a problem with the arrays and that's why Illumina discontinued them!! I am (*&^%£ off because Illumina assured us they were ok...but one of the controls on the array has NEVER worked (!"!!).

Anyway, as we had 12 more arrays left, I have submitted more samples to see if we get the date effect again (or perhaps our pilot experiment was just rubbish).

Clare

Hi,

I haven't run Illumina's MicroRNA chips but I have performed profiling using the GEX chips, where we have performed several pilot projects at different time points. If we wanted to be able to compare results from two different runs, then we would usually include a few of the samples from a previous run along with the present run...just to see if there is any technical variation. I'm not sure if you have already tried this, but it could be worth a shot.

Good luck!

[onclab]

Clare, on Jun 10 2010, 10:53 AM, said:

We ran another 12 samples on the same platform, including samples that were run on the first 2 experiments. Samples that were run on both the 2nd and 3rd experiment cluster together nicely - so for some reason the pilot experiment we did just didn't work

Clare

onclab, on Jun 9 2010, 07:15 PM, said:

Clare, on Jun 2 2010, 12:51 PM, said:

Yes, date clustering sad.gif We now think it's a problem with the arrays and that's why Illumina discontinued them!! I am (*&^%£ off because Illumina assured us they were ok...but one of the controls on the array has NEVER worked (!"!!).

Anyway, as we had 12 more arrays left, I have submitted more samples to see if we get the date effect again (or perhaps our pilot experiment was just rubbish).

Clare

Hi,

I haven't run Illumina's MicroRNA chips but I have performed profiling using the GEX chips, where we have performed several pilot projects at different time points. If we wanted to be able to compare results from two different runs, then we would usually include a few of the samples from a previous run along with the present run...just to see if there is any technical variation. I'm not sure if you have already tried this, but it could be worth a shot.

Good luck!

When running microarrays, degradation of the dyes used is a problem at certain times during the year when ozone levels are higher (for example, during the warmer months of the year). If all of your arrays were run during the same time of year, then hopefully this won't be an issue.

Here's some info on the effect of ozone on flourescent dyes, Degradation of dyes