Viability at a given dosage from AUC

For PRISM 19Q4, secondary screen, I am wondering if we can compute % killing at a required dosage from AUC values. For each drug-cell line pair, we have available fitted curve upper/lower limits & slope.

If I am not mistaken, these two formulas from Corsello et al. 2020 could be the starting points, but am not sure about the next steps as not all variables (eg. E-inf) here are completely clear. Thanks! Screen Shot 2020-12-26 at 10.54.22 PM
Screen Shot 2020-12-26 at 10.57.41 PM


I believe your comment is totally correct, we can essentially “impute” the missing doses using the dose-response parameters we computed.

To do that you should download “secondary-dose-response-curve-parameters.csv” file and use “lower_limit” column as the E_infty parameter in the equations.

However, for a simpler solution one can just take replicate level log-fold change data and fit a line using the available data for the closest 2 doses (one above and one below the targeted dose - in total ~6 data points: 2 doses x 3 replicates) to approximate the final result. This will help you with the missing dose response parameters (i.e. poor dose response curve fits), and would be easier to track.