Hi,
It is a wonderful project and I am personally trying to make the reward function that you mentioned in your paper using Self-Organising Maps (SOMs).
For this I used both Morgan Fingerprint and Maccs Fingerprint as features. But the clusters in SOM are not converging and are scattered in
I also tried with features like Molecular weight, Synthetic accessibility, LogP, number of rings, QED score and Natural Product likeness. But still it did not converge.
So, I wanted to ask, what features did you use for creating these SOMs. I just want to create Kinase SOM and Specific DDR1 SOM.
The (b) and (c) in this picture:
Hi, It is a wonderful project and I am personally trying to make the reward function that you mentioned in your paper using Self-Organising Maps (SOMs).
For this I used both Morgan Fingerprint and Maccs Fingerprint as features. But the clusters in SOM are not converging and are scattered in
I also tried with features like Molecular weight, Synthetic accessibility, LogP, number of rings, QED score and Natural Product likeness. But still it did not converge.![SOM_kinVnkin_1000_100x100](https://user-images.githubusercontent.com/11424274/66540473-8d766f80-eb49-11e9-9f33-b8a57841e520.png)
So, I wanted to ask, what features did you use for creating these SOMs. I just want to create Kinase SOM and Specific DDR1 SOM. The (b) and (c) in this picture:![SOM_reference](https://media.springernature.com/full/springer-static/esm/art%3A10.1038%2Fs41587-019-0224-x/MediaObjects/41587_2019_224_Fig4_ESM.jpg?as=webp)