Every complex has predicted pKa of 6.041

[allowishus]

Hi, thanks for this research! I've been playing with the trained model and attempting to reproduce some of the paper's results via the readme instructions. I followed the build, data preparation, and feature generation steps without issue. For some reason each predicted pKa comes out to exactly 6.041.

predicted_pKa.csv output excerpt:

/pdbbind_samples/9hvp_cplx.pdb,**6.041**
/pdbbind_samples/9icd_cplx.pdb,**6.041**
/pdbbind_samples/9lpr_cplx.pdb,**6.041**
...

Have you seen anything like this before or can you point me in the right direction as to the possible root cause?

[zhenglz]

Thanks for using the model. I encountered the same problem before, possibly due to inappropriate protein-ligand complex preparation. Please send me the complex pdb structure and related feature file, I try to duplicate the error and find a solution. My email is: astrozheng_AT_gmail.com

[allowishus]

Thanks for your help! I emailed you the data.