@Florian Thank you for the suggestion. I'm not sure if there is a way to view the raw data, this software seems a little lite on features but I may just be inexperienced. From my view I see three options for viewing the data, "Background Subtracted", "PCR Baseline Subtracted", and "PCR Baseline Subtracted Curve Fit", with the first being the closest to raw data.

@Danail Thank you very much for your idea, I think you may be on the right track here, although from my understanding it seems that the MyIQ is a very basic instrument which only measures a single optical channel, so I believe fluorescein is spiked into the supermix and measured on the same channel as the SYBR green.

I have noticed that this instrument supposedly does not support low-profile PCR plates, so there is a large gap between the top of the heating block and the lid heater. Looking at plates or tubes after a run, it doesn't seem like a noticable amount of sample has completely evaporated, but perhaps it is temporarily accumulating on these un-heated sidewalls and throwing off the signal.

I have some more tests I'm trying to do to figure this out, I will post here again with more info in the next 24 hours or so.

Florian C Priller

I have found that running a program on wells containing just dye (External Well Factor Calibrator solution) also produces weird results with large 'steps' in the signal, and it does seem like evaporation is occurring, not sample loss, but condensation on the side wall. As I think you are saying, it seems that the instrument only supports high (maybe the same as 'regular') profile plates (~20.7 mm) tall, which leaves a large gap between the heatblock and the top of the plate or tube where condensation can collect. I have samples of ~15 mm tall plates and these do not seem to make contact with the lid.

The thing I still don't know is why the problem seemed to crop up all of a sudden despite using the same plates as always.

I am currently running an experiment using different plates which have a skirt, and testing the effect of 20 vs 50 uL reaction volume and with our without mineral oil overlay to see if these will get rid of the behavior.

I have monitored the data output during the run and assuming that is raw data, I can tell you that the decreases in signal in the second half of the run are also visible there.

I am looking into getting a Bio-Rad representative out to look at it, I will post here if I figure it out or when I have more information for anyone else who may find themselves with the same issue in the future.

Thanks again Florian.

The phenomenon of amplification curves that tilt downward is often caused by an improper baseline correction. In some cases, the fluorescence will drop more than a small amount. If the automatic baseline correction starts to calculate the baseline in cycles that show still dropping fluorescence, the baseline will go downward. Herewith too much fluorescence will be subtracted from the actual fluorescence that is generated by your qPCR. Please look at non-baseline corrected data to see if this happened. And if so, correct the baseline manually and set the cycle to measure the baseline at a higher cycle number. Please do not use such amounts of targets that you will find Cq's below, say 15, Otherwise the calculation of the baseline will be difficult.

Florian C Priller

I have been sealing plates with optical sealing film and there has been no visible sample loss and the seal is still well affixed after the run. The problem also seems to occur when using 8-well strips with optical flat caps, but I have only tried that once I believe. I tried to check the remaining volume by pipette in the latest run and it was hard to determine accurately, but it seemed that there might be only around 15 uL volume remaining from 20 uL wells after 60 cycle (longer run for troubleshooting purposes). There was not a drastic or total loss though.

I have some new data which I think should prove useful, but I am not sure how to interpret it. In this experiment I used a new type of skirted plate which is much stiffer, again using optical sealing film.

AmpCurve1.jpg shows duplicate wells containing either 20 uL total volume (Green, 4000 RFU peak) or 50 uL total volume (Purple, 8000 RFU peak).

The concentration of input (2.5 ng cDNA per 20 uL), primers (300 nM each), and master mix are equivalent. The target here is mouse Hprt, a 132 bp amplicon.

I also prepared a matching set of these wells overlaid with 10 uL each of Bio-Rad mineral oil, with the plate being centrifuged briefly up to 800xg before the run. These oil-covered are shown in AmpCurve2.jpg, the 50 uL wells are the two which hit their peak slightly earlier.

The addition of oil seemed to significantly improve the signal peak in the pair of 20 uL wells as well as providing a steeper upwards slope (higher PCR efficiency?), but the severe drop in signal still occurred. With or without mineral oil, 20 uL wells seemed to show an obvious step in signal around cycle 47, whereas 50 uL wells did not show this step but dropped to their lowest value sooner around cycle 37.

AmpCurve3 shows single replicate controls, yellow is 20 uL reaction without cDNA input (0 RFU initial value), green is 20 uL of water (-600 RFU initial), cyan is 50 uL of reaction without cDNA (+1700 RFU initial), red is 50 uL of water (+400 RFU initial).

As you can see, the drop in signal starting around cycle 31 and concluding around cycle 38 happened in all of these wells, normal reactions, increased volume reactions, reactions with mineral oil, reactions without cDNA, and even reactions with just water.

I also prepared n=4 replicates containing the Bio-Rad External Well Factor solution, which I believe contains fluorescein but I don't know. It is pinkish to the eye and glows orange under UV. Again the signal drop occurred in a similar region. See AmpCurve4. Wells with 20 uL 1x dye start around 600-700 RFU, wells with 50 uL of 1x dye start around 900-1000 RFU. All of these wells suffer loss of signal in the 30-40 cycle region.

The melt curve program run immediately afterwards showed no signal and no peaks, all wells showed a starting and ending signal (basically flat) below 0 RFU.

Then the lamp was turned off and the samples sat in the machine until the next day and the melt curve was re-run (1 min at 95 deg C followed by 50-95 deg C by 0.2 deg C steps, 8 sec dwell time).

At the start of this melt curve, wells containing dye or reactions were showing RFU's above zero, and well-defined melt peaks could be seen in the positive control wells. The highest RFU values observed matched those from mid-run yesterday at around 7000 RFU.

I don't know that you have the time or patience to read this incredibly long explanation, but I thank you for your help nonetheless.

Hopefully this establishes that the problem has nothing to do with DNA degradation, and probably has nothing to do with loss of SYBR or fluorescein either if signal drop can be observed in wells containing just water. I wish I had access to raw data or better understood what kind of corrections this instrument performs on the data. Since increasing reaction volume to 50 uL or adding mineral oil doesn't seem to solve the problem, it seems unlikely to be due to evaporation either.

Lamp failure seems to make sense, but it is strange that the signal looks stable in the first 20 cycles and that running a 90 minute melt curve the next day gives no sign of problems. I have a new lamp coming in the mail on Monday from eBay, Bio-Rad has discontinued it. In the mean time I'm about ready to give up on this and find another machine to use, but that will be a lot less convenient.

Thanks again.

In case anyone finds this question looking for an update, I did not ever conclusively identify the issue, but a Bio-Rad technician looked over the instrument data files and concurred that it was probably a failure in the detection hardware, most likely the CCD itself. Since the MyIQ is past its end-of-life for first-party service, we have not made any attempt to get it serviced.