Dim serum signal - 20200417_132123_311 - B12-0004

[tischi]

{"plateName":"20200417_132123_311","siteName":"B12-0004","pixelLocation":[35294.45972158684,4694.54366210524,0.0]}

@imagirom @metavibor

In this well the serum staining is overall very dim. In fact it is so close to the background noise that we are for some cells getting negative numbers after the background subtraction.

In such assays one usually also has an automated QC when the signal that one wants to measure is too close to the background noise, because then the measurement becomes unreliable.

• [ ] Should we introduce this? We would have to think how to do it exactly.... I think something like the signal should be brighter (in the control cells) than plate_median_bg + N * plate_mad. Otherwise we discard this image.

Screenshot

[metavibor]

@tischi @imagirom I would not discard those... it can happen that there is nothing in the serum that binds to cells. But we should mark these wells somehow so we can potentially check them one more time... I'm thinking of a potential pipetting error where no serum or very little serum was added to this well

[tischi]

Here is are some plots of the variation of the intensity of the control cells.

The intensity differences are quite significant:

Importantly, I do not see an immediate correlation with the ratio:

@metavibor What could we write in the publication about the variation of the intensity of the control cells? Is there some sample prep step that could cause this or is this some biology? [EDIT: You just answered... 😃 ]

[metavibor]

one way to remove the unspecific background that sticks to control cells would be to "pre-adsorb" the antibodies on non infected cells. One basically adds serum on regular cells for some time, collects the supernatant and adds it to the wells of mixed infected and non infected cells. This would clean up the serum from "sticky" antibodies and would probably make the assay better. But we decided against it for now, it adds an additional step which was too much to handle manually. Once we move to fully automated liquid handling than this should be done.

[tischi]

...to "pre-adsorb" the antibodies on non infected cells...

good idea, but it would not solve our issue but make it worse, because the control cells would become even dimmer. We would have to change the score to something where we do not divide by the control cell intensity (which would become a very unstable potentially zero measurement).

I would not discard those...

We would automatically mark them as "dim serum", so you could always go back to them and redo the experiment. But I think we should exclude them in a sense that it was not possible to obtain any reliable measurement from them in this plate.

[metavibor]

in other words - having some background in control cells is welcomed. Having too good assay where the control cells become almost "black" would be a problem

But I think we should exclude them in a sense that it was not possible to obtain any reliable measurement from them in this plate.

I would not say that any reliable measure can not be obtained, It depends what is the ratio in that well... In case of a very high ratio which can happen by diving with a very low number I would manually check the well and see if I can visually assess it. In case when the ratio is 1 than I would say it is as reliable as other "1"s

[tischi]

in other words - having some background in control cells is welcomed. Having too good assay where the control cells become almost "black" would be a problem

Essentially, I think we need a "pipetting control" for each well, because otherwise we cannot know whether a well is dim or bright because of better binding of IgX to the virus or whether the pipette was more or less blocked. And the only idea so far is the binding of the serum to control cells. However also that is not ideal because it may contain some biological information, as we discussed. I am still a bit confused and have the feeling that there should be something better but I don't have an idea immediately. What do you think?