advice: structure search improvment

Dear support,

 

Actually I upgraded jchem from 5.1.4 to 5.2.6 and thought that search performance should have been greater. I made several benchmark, and I don't find that jchem search has improvment in speed. The problem might be because we have the samed structure table since the beginning.

We have the size of fingerprint which is 32 (numberOfBits = 1024), numberOfOnes=2 and numberOfEdges=6. We store only molecules, but actually tableType is set to 2.

Do you think that it can improve noticeably, that I change the structure table to greater size of fingerprint, Ones and Edges?

Thank you in advance for your advices and help.

Warmest regards,

Jack

Hi Jack,

Indeed, the average speedup in version 5.2 is due to increased screening efficiency, resulting from changes in the original algorithm. However, the magnitude of the speedup may differ depending on the database content, the individual queries used and fingerprint settings. (In fact, there are queries which may result in slight slowdown as well.)

The published average speedup benchmarks were recorded using the default fingerprint sizes. (512,6,2).

Your longer fingerprint size (1024 length) seems to give quite an efficient screening for your database content and test queries already in version 5.1.

 

Best regards,

Szabolcs

Hi Szabolcs,

Thank you so far. Your explanations are clear.

Kind Regards,

Jack