meyer-lab / cell-selective-ligands Goto Github PK

It seems as though the sigma point method handles R1hi R2hi strangely when compared to other populations (namely R1hi R2lo and R1lo R2hi). This is evidenced by the optimization figure (figure 6) where greater selectivity is achieved for R1lo R2hi than R1hi R2hi (shouldn't be possible) as well as the new handling of the population to population comparisons (fig 4) where some strangeness was occurring.

The axis labels are missing in (a). Remove the text on the very top of the figure.

Try making a bar graph instead of heatmap for b–e. Also, I think it'd help to have b/c/d on the same scale.

In the figure caption, the text reads a lot like a results section, not a caption. The caption should focus more on explaining the elements of the figure, not their interpretation.

I still feel like a nice way to end the manuscript would be to make something like this, where we optimize for the best result for each cell population:

This would help to reinforce that (1) different strategies are helpful in different situations, (2) the amount of selectivity attainable is different for each situation, and (3) combining strategies does/doesn't lead to synergistic benefits.

Can treat cell populations as Gaussian populations with, say, five receptors:

X = [x1, x2, x3, x4, x5]

We can assume the stdev of the populations is 1 throughout.

Population 1 could be centered at the origin: v = [0, 0, 0, 0, 0]
Population 2: v = [1, 0, 0, 0, 0]
Population 3: v = [0, 1, 1, 0, 0]
Population 4: v = [10, 0, 0, 0, 0]

We could define selectivity as Lbind_target / sum(Lbind). Can reformulate this on the full real axis with logit function.

It doesn't have a doi/pmid since it's someone's PhD dissertation.

You both should make ORCID IDs, and then add them here:

https://github.com/meyer-lab/cell-selective-ligands/blob/master/manuscript/metadata.yaml

https://orcid.org

Probably just sample from a distribution some set # of times?

Some of the output calculations are extremely skewed. I'll change the sampling to report 10/90% quantiles, but then I'll need help updating the plotting.

I guess we can consider both untethered mixtures, and hetero-oligomeric molecules.