Cluster t-SNE plot [message #500] |
Sat, 16 March 2019 07:28 ![Go to next message Go to next message](/forum/theme/myThemeWithHFCSS/images/down.png) |
amorrison
Messages: 36 Registered: March 2016
|
Member |
|
|
Hi,
I'm really liking the t-SNE visualisation. If I wanted to bin the compounds based on their t-SNE positions is this possible and how would I go about it? So similar to the cluster compounds but using t-SNE to a certain distance.
Thanks,
Angus
|
|
|
Re: Cluster t-SNE plot [message #506 is a reply to message #500] |
Wed, 20 March 2019 23:29 ![Go to previous message Go to previous message](/forum/theme/myThemeWithHFCSS/images/up.png) |
thomas
Messages: 696 Registered: June 2014
|
Senior Member |
|
|
Hi Angus,
this is currently not possible. I would suggest to always cluster compounds on a chemical descriptor, because that would reflect compound similarity much better than the reduced 2 or three dimensions after t-SNE. Of course, clusters based on the t-SNE positions would look perfectly nice in the t-SNE view, but would be far from perfect groups of similar compounds.
Thomas
|
|
|