Similarity analysis using "find similar compounds..." - slow analysis of libraries [message #1133] |
Tue, 24 November 2020 12:35 |
SM2020
Messages: 13 Registered: June 2020 Location: UK
|
Junior Member |
|
|
Hi,
I'm carrying out a similarity analysis by comparing x2 cpd libraries (each ~ 16K) using the "find similar compounds in file..." option using the FragFP descriptor. I am selecting a similarity limit cut-off of ~30% (using the slide bar selector).
I want to obtain the nearest neighbour measure for a given library, so nearest neighbour values can be binned and plotted (note: I realise there are other methods for library comparison as well in DW).
The process is taking an extremely long time on my (admittedly) old computer (4-core/8 threads, i7-3615QM, 16GB Ram, MacOS) - > 24 hrs.
With that in mind, a couple of general questions...
1) What improvements (if any) would help in speeding up this process?
Would increasing the accessible RAM help, or am I generally limited by processor speeds for such analyses?
2) What general hardware upgrades and/or software upgrades would you suggest for dramatically speeding up these types of analyses (<< 24 hrs) in DW?
Faster processor/s with more cores/threads? More RAM? Both...? Other...?
Best Wishes and many thanks.
|
|
|