G1v2 Configuration and Results

[eric-czech]

Hey @jmotschman ,

This includes a configuration for the G1v2 chip. Results from processing the related data look like what I imagine you'd expect:

Growth rate distributions

Average Cell Counts

Growth Patterns cc: @benjaminyellen

This was an exploration based on https://github.com/hammerlab/celldom/issues/59 where it's pretty clear that linear growth models (in log space) aren't capturing a lot of what's going on with treated apartments. Below are all of the cell count timeseries for each apartment (in both arrays in this case) projected into a 2D space where each point represents a single time series and the proximity of the points represents distance on a nonlinear manifold that models differences in growth patterns:

I think the purple group is particularly problematic because it's 20% of all apartments, does not fit a linear model well, and I think best explains why there is that initial divergence in the predicted vs actual counts in the Average Cell Counts graph above.

I think it may be worth looking at something like this regularly as a way to make sure whatever growth model we use isn't breaking down like this too often. Anyways if that kind of analysis makes sense to you and you see some ongoing value in it, I'd be happy to make it another template notebook command or to potentially figure out how to add it to the app.

---

For the sake of reference, the new chip config is at chip-G1v2.yaml and the experiment config at exp-20181013-G1v2-K562-imatinib.yaml

[jmotschman]

Thanks @eric-czech . I was able to get this to work here on our machine. The controls look similar as well.

G3v1 and G1v2 below: