Known starting structure

[domgurvic]

Hi Patrick,

Thanks for creating and sharing EvoBind, I had it running on few different proteins and the results are pretty interesting.

I was wondering if there is a way to manually select the starting composition of designed peptide that I want to improve, rather than starting from a random sequence?

Thanks!

Dom.

[patrickbryant1]

Hi,

Great to hear you like it!

Yes, you can write a metrics file with the starting sequence (run predict only) and then start from that or simply enter it in colab.

Best,

Patrick

[domgurvic]

Thanks for getting back to me so quick!

I changed --predict_only=True and added --peptide_sequence="ACSFPKCGCA" for the test case of 1ssc that you provided.

But I can't seem to bypass it making a prediction but plddt score being zero and resulting in this error:

` loss = if_dist_peptide*1/plddt

ZeroDivisionError: division by zero
` 

Also since it made a prediction and saved a structure unrelaxed_1.pdb I can see that the starting structure is not the one I provided with a flag but a random one. 

Any advice?

Best,

Dom.

[patrickbryant1]

Hi,

There was a bug - this should be fixed now. Please do a pull and try again.

Best,

Patrick

[domgurvic]

Hi,

You're right, its fixed now and it works exactly as you suggested with the starting structure.

Thank you.

I'm happy to close this down but i do have another question regarding starting structure and distance to the target residues, is it best to open another thread?

Dom.