Create node to test multiple thresholds for ID spots in Cellprofiler pipeline?

[Lois]

Hi all!

I am new to Knime. I have turned to Knime because I want to reduce variability between experiments by choosing the best threshold for each experiment. Testing multiple thresholds for 7 samples in 5 experiments is time consuming in Cellprofiler and I am hoping to automate testing thresholds in Knime.

Looking at the “speckles” example workflow, I think what I want to do is create a new node before the “Execute pipeline” node that lists multiple thresholds to test. As I said, I am new to Knime. Is this how you would go about it @christian.dietz ? Has anyone already made a similar workflow in Knime? I think it would be generally useful. I have colleagues who count spots by hand instead of using automated image analysis because of the problem/time-consuming process of selecting a threshold for each experiment.

For some more background on my setup: I have a Cellprofiler pipeline that detects nuclei, red spots and green spots. It then assigns each spot to a nucleus and counts the number per nucleus. Finally, it measures the number of overlapping spots per nucleus.

Cheers,
Lois

[stelfrich]

Hi @Lois,

Looking at the “speckles” example workflow, I think what I want to do is create a new node before the “Execute pipeline” node that lists multiple thresholds to test.

You are right that creating a list of thresholds could be a first step to approach this task. Once you have generated such a list, you can execute parts of a workflow each time using a different threshold using loops: Before trying to tackle your task, you might want to read up on the concepts of loops and flow variables.

This, however, does currently not work with CellProfiler pipelines, since they are not “parameterized”. That is, we can’t change parameters of the pipeline when executing it. In order for the aforementioned approach to work, you’d have to migrate your CP pipeline to a KNIME workflow: If you are open to that, I can dig out an example workflow that does the kind of colocalization analysis that you are interested in…

Best,
Stefan