search

hci-unihd / antibodies-analysis-issues

Issue tracker for problems in the antibodies analysis workflow.
0 stars 0 forks source link

Cross contamination IgA/marker? #47

Open lorenzocerrone opened 4 years ago

lorenzocerrone commented 4 years ago

plate_name: plateK12rep1_20200430_155932_313_IgA file_name: WellE12_PointE12_0006_ChannelDAPI,WF_GFP,TRITC,WF_Cy5,DIA_Seq0537

Not sure what is the exact issue here, @tischi do you have an idea? is cross contamination the right term?

Screenshot_2020-05-02_18-22-43

constantinpape commented 4 years ago

Have you seen this before, @metavibor?

tischi commented 4 years ago

@lorenzocerrone could you give a bit more detail?

  1. I assume this is the virus channel, correct?
  2. What's the issue with this image in your opinion?
lorenzocerrone commented 4 years ago

@lorenzocerrone could you give a bit more detail?

1. I assume this is the virus channel, correct?

2. What's the issue with this image in your opinion?
  1. No this is the IgG channel, the virus channel looks alright.
  2. I am not sure. It's the first time I see an image looking like this so I thought to share it. To me it looks like there is some sort of "contamination" between the IgG and marker channel, but I really do not have more thoughts. I guess that from the analysis point of view those should be marked as outlier, right?
tischi commented 4 years ago

@lorenzocerrone Very good of you to post it here!

  1. Did you look at the images next to this one in the same well? Do they have the same issue?
  2. Are you sure this is IgG and not something else, like IgA?
  3. Could you also post the virus channel next to the IgG channel such that we can see whether the signals look similar?

One important thing on @metavibor TODO list is already to experimentally assess whether we have some channel crosstalk. So we should find out.

constantinpape commented 4 years ago 
2\. Are you sure this is IgG and not something else, like IgA?

Also IgA shouldn't look like this, it really looks like some cross talk from the virus channel. (Of course we should still know which one it is).

3\. Could you also post the virus channel next to the IgG channel such that we can see whether the signals look similar?

Yeah, that would be goog!

lorenzocerrone commented 4 years ago 
2. Are you sure this is IgG and not something else, like IgA?

My bad @tischi, I confused IgA with IgG. The image posted is from IgA 100%. I will have look at the whole plate now and let you know about 1. 3.

metavibor commented 4 years ago

@constantinpape @lorenzocerrone nicely spotted guys! @tischi I have seen these and reported on them when we had meeting with Wolfgang. This is definitely not a crosstalk from the virus channel. But - yes it looks totally like the virus channel. So what is it? I have seen this in cca 10 different wells with IgA so far (never with IgG)... I have a crazy theory about it but I will leave this when we meet via zoom. For this reason I don't consider IgA alone to be reliable marker for SARS-CoV-2 antibody detection (several negative controls have shown this type of IgA staining pattern). When there is a positive hit with IgA I always check the images manually and if I see this staining pattern I consider it highly likely to be false positive. That is why I think we cannot rely solely on ratios to call positive hits. In this case there will be quite high ratio. This is where "machine learning" comes into play for calling positive hits because not only ratio but a staining pattern needs to be concerned

lorenzocerrone commented 4 years ago

@metavibor thanks for the clarification, I will keep my eyes open if I spot something similar on IgG.

@tischi @constantinpape 

1. Did you look at the images next to this one in the same well? Do they have the same issue?

I checked with the plate viewer (works great!!) on the whole plate and I found the same "effect' in the following locations: E12 (the one from the screenshot), B04, F02, F11