On Feb 26, 2009, at 5:32 PM, Igor Filippov [Contr] wrote:
Interesting, but 90,000 structures is a tiny database by today's
standards.
I saw that and agreed. Brute force that's about 10 seconds with OEChem.
Another point worth
mentioning is that similarity search is not the same as substructure
search and it seems like Duan is equaling the two.
Specifically he writes:
50ms to do substructure search on chemsoso.com’s 90,000 structures
while
For similar structure search, it takes about 300ms
I've been talking to people to get an idea of fingerprint filter rates.
It seems 85-90% is a good estimate of how many compounds are screened
out. These numbers don't agree with this being a full substructure
search.
On Thu, 2009-02-26 at 15:47 +0000, baoilleach wrote:
Database searching using fingerprints seems to be a topic of interest
at the moment...
And I'll be writing my experiences up during the next week
for my OpenEye CUP presentation "Database Extensions for fun and
profit."
:)
Andrew
[email protected]
