Optimizing Global Similarity Score Monitoring in Reinvent with Transfer Learning

[natarajbalakrishnan]

We are using transfer learning techniques based on Reinvent. Our goal is to monitor the global similarity score of all training SMILES versus the SMILES generated in each epoch. There is a way to do this using the RDKit bulk similarity search routine outside of the Reinvent package, but it hinders the performance of the workflow. Therefore, we would like to hear from you about the most efficient way to achieve this within the Reinvent package. Your help would be greatly appreciated.

[halx]

Hi,

many thanks for your interested in REINVENT and welcome to the community!

Incidentally, I am currently working on this with a collaborator. But it will take some while until this will be ready.

Many thanks,
Hannes.

[halx]

Just to clarify: this is only about monitoring not scoring?

[natarajbalakrishnan]

Hi thanks for your reply, yes just for monitoring

[halx]

It has been a while but now have a solution for this with iSIM implemented by the Miranda Quintana Group at UF.

[natarajmtech]

Thanks for your update on this feature availability ,it would be helpful for our research