Fig. 7: Testing of the control condition-trained model on different cell conditions.

Testing of the model trained by data in control condition on other cell conditions. a Test on control condition (condition 1). b Test on quiescent cell data (condition 2). c Test on data from cells treated with 50 μM Bleomycin (condition 3). d Test on data from cells treated with 200 nM Doxorubycin (condition 4). In all the results, the model trained by data in control condition provides satisfactory accuracy compared to the full neural network and this means that the intrinsic mapping between standardized input and standardized P53 content remains invariant across different cell conditions. The only difference among different conditions is the probability distribution.