variation of predicted pIC50 for the same compound, among several running trials #4
Description
Hello,
I installed CardioGenAI on a Linux machine using the provided "env-linux.yml" (env-flex.txt).
I have tried evaluating the pIC50 of a compound several times (using the attached script and utilizing either GPU or CPU)
predict_cardiac_ion_channel_activity.txt.
However, I observe noticeable variations among different trials for the same compound. Any idea what the issue might be?
Thank you
predictions_cpu_1.json
predictions_cpu_2.json
predictions_gpu_1.json
predictions_gpu_2.json