More Multi-Spectral qPCR Experiments
In a previous post I discussed some experiments replacing much of the complexity around the detection optics in qPCR systems with cheap multispectral sensors.
To briefly summarize: You can get these cheap $10 sensors which allow you to sense 10 different wavelengths (or colors if you like) in a compact package. CoDx mention these for use in qPCR in their patents. So I hacked one into an old qPCR machine to see how well it would work.
I previously left off just before running any actual qPCR experiments. After some false starts1, detecting with the multi-spectral sensor seems to work fine. Here’s the output of a positive control from a qPCR kit I’ve previously used:
And here the same sample on a regular SmartCycler:
The regular SmartCycler output is obviously better, but both show a CT of somewhere around 20. I’m yet to optimize the system. In particular, the multi-spectral output is from a single time point (where I could be averaging over multiple). I’m also not doing anything in terms of background correction (which the SmartCycler does). I’ve yet to play with the integration time at all. And of course the whole thing is only roughly aligned and held together with epoxy.
But results are good enough to give some confidence that the approach works in my view.
Next I’m wondering if the excitation side might also be simplified. Not much use removing all the filters from the detection side if you still need to clean up the LED based excitation. Either “better LEDs” or lasers seem like good options here?
In my first run I used the default gain settings. This produced some quite odd results:
I figured perhaps the qPCR machine was for some reason slowly increase the excitation with cycle and I was just seeing some of the excitation bleeding through (perhaps due to poor filters). I tried correcting for this in a number of ways… but nothing worked. I was using the positions identified in the previous post to extract the fluorescence. But looked elsewhere in the trace too…
Then today I tried another run with the sensor gain set of max and everything seemed to work fine. The sensor does have automatic gain control, perhaps setting it to max gain disables this? But in any case I suspect the above trace was an artifact of the AGC…