a bunch of healthy organoids disintegrated due to mechanical errors today. and i was spiraling for a good amount of time (4 hours?).
everything holds true that as we move toward day 120, fewer samples will survive, and perhaps we will reach below the ENR threshold (n<50).
and then the moment i realized that foxg1 becomes bimodally distributed... actually all the surviving ones should have it, which means its universally 1, or skewed left.
the correlation of the varying morphology and a stagnantly bimodal or skewed left will not correlate to each other if the model isn't non-linear, which it isn't. ENR is linear.
non-linear models are capable of solving everything ENR can't simply solve by interpreting linear relationships, but that's a future question, given funding, given scaling data.
but right now, what is the best window for the sake of normal distribution of ground truth aka biggest divergence aka largest positive derivative of the expression level curve of key biomarkers for cortical fate (differentiation) as well as maturation, and also considering the fact that we need enough sample size in the first place?
the answer is... by the end of this week.
there's more to confirm (specifically about what biomarker, based on our very big variation of culturing protocol compared to actual pasca, not just identical in feeding schedule but all environmental conditions, in which ours is stationary in incubator not on shakers or bioreactors) with sun very soon, but this experiment will see the light of day, much faster, with much higher potential of success, thus more ideal than the original ideal, where every biomarker's expression level by day 120+ skews left at right end, not normally distributed. but we need normal distribution for ENR.
this is perhaps the single biggest breakthrough i've ever done, all because i thought i was cooked.
im really really really tired right now. but at least i can sleep peacefully for tonight.