Selecting negative and positive control ORFs

[MarziehHaghighi]

Negative controls:

• We want negative controls to:
  • Have no signature confirmed by existing datasets.
    • Having no signature usually is equivalent to having low replicate correlations for negative controls, so relying on that rule, Niranj has searched for negative controls which have low replicate correlations based on two CPJUMP1 and in the JUMP production experiments.

Positive controls:

• We want positive controls to:
  • Have strong wt to mutant mislocalization
    • we check the manual impact score by Jessie
  • Have strong wt phenotype using the rest of the cell painting channels comparing to controls
    • we check the replicate correlation in the cpjump1 dataset.

[MarziehHaghighi]

Source Notebook

Negative control candidates:

• PRKACB, RB1 , MAPK9, RHEB seem to have low rep cor in cpjump1 and relatively smaller rep cor in jump production

Positive control candidates:

• None of the genes have high replicate correlation according to cpjump1 but if we want select just based on this overlap we can pick IMPDH1 and ALK

[AnneCarpenter]

Neg controls sound great to me! Will you let the Taipale lab folks know and cc me?

For positives, there was another criterion on the expt planning doc for the triplet we want to choose: 1) WT + two mutants, one subtle/one strong in the protein channel 2) AND Cell Painting morph changes (in other words, we wanted a single gene with two mutants that met all these criteria!)

It may not be possible to find such a triplet systematically - it relies on Taipale lab folks thinking of candidates based on (1), and then checking for (2) in our data.

A triplet like this may not even exist, though, so if Taipale lab doesn't have any candidates (for which we also have Cell Painting data to test criterion 2), then the best we could achieve is perhaps: A. One WT-MUT pair with obvious localization change (Taipale lab can just choose one they like) B. One WT gene that gives a strong morphology signal in Cell Painting (eg from CPJUMP1) To give us 3 constructs total. Perhaps ask Taipale lab which pair they prefer for A and then just pick something for B from CPJUMP1 data?

[MarziehHaghighi]

Jessica's note:

Based on the list you provided I think there are two good options:

An IMPDH1 triplet: IMPDH1 WT - IMPDH1 D311N (low/mid confidence 3) - IMPDH1 R309P (high confidence 5)

Or: ALK + ALK variant + WT of your choice from CPJUMP1

I can ask Chloe to do a pilot with them to make sure they're still robust in our new system (by eye for now).

Let me know what you think,

[AnneCarpenter]

It sounds like IMPDH1 does not have a strong replicate correlation in CPJUMP1? @MarziehHaghighi If so then we better choose the second set here, ALK + ALK variant + WT of your choice from CPJUMP1

[MarziehHaghighi]

It sounds like IMPDH1 does not have a strong replicate correlation in CPJUMP1? @MarziehHaghighi If so then we better choose the second set here, ALK + ALK variant + WT of your choice from CPJUMP1

@AnneCarpenter hmm they both seem to have ~0.2 replicate correlation according to positive control plot?

[AnneCarpenter]

I thought 0.2 is pretty low, in absolute terms (even though it's the highest of the set of samples you plotted). Typically we like to see above ~0.35 to be confident a sample has a strong signal - then again I've no idea if the normalization done for this dataset is similar to our usual and obviously 0.35 is a rule of thumb. But I would say unless you have reason to believe that 0.2 counts as a good signal in this dataset ,I would go with the second set here, ALK + ALK variant + WT of your choice from CPJUMP1.

[MarziehHaghighi]

@AnneCarpenter the point I was trying to make was that both of "ALK" and "IMPDH1" have the same replicate correlation values according to the CPJUMP1 data. Am I missing anything?

[AnneCarpenter]

My recollection is that we are choosing positive controls here for two reasons: localization changes and cell morphology changes. "ALK" and "IMPDH1" will let us see localization changes but it sounds like neither will be strong on cell morphology changes (because both are ~0.2). This means we need to go with option 2 from her note

I don't know why jessica originally suggested these two options: An IMPDH1 triplet: IMPDH1 WT - IMPDH1 D311N (low/mid confidence 3) - IMPDH1 R309P (high confidence 5) Or: ALK + ALK variant + WT of your choice from CPJUMP1

Instead of proposing: iIMPDH1 WT + IMPDH1 R309P + WT of your choice from CPJUMP1

But since she seems to prefer ALK + ALK variant + WT of your choice from CPJUMP1 then I think we should go that route

[MarziehHaghighi]

I see! @niranjchandrasekaran do you have any favorite impactful perturbation in CPJUMP1 set?

[niranjchandrasekaran]

I don't have any favorites, as I didn't focus on individual genes in my analyses. But if we assume replicate correlation correlates with impact, then the genes at the top of the replicate correlation list would be the ones to go with.

[AnneCarpenter]

Great - I'm a bit nervous that DYRK1B is so much higher than the others (sometimes when things are too good it's a technical artifact), so unless someone wants to go check those images, I think I might suggest PTK2B - its function is very plausible to cause morphology changes.