fragment-based-dgm icon indicating copy to clipboard operation
fragment-based-dgm copied to clipboard

Published 20 hours ago • verified publisher icon marcopodda

Why not directly use a simple RNN to generate molecules (like REINVENT) ?

Open jkwang93 opened this issue 3 years ago • 3 comments

Hello~ I have a new question to ask you😂. Why not directly use an RNN to generate molecules (just like REINVENT), but use a seq2seq model?

jkwang93 avatar Apr 12 '21 11:04 jkwang93

Not sure what you mean, this is simply a different approach to the learning task.

marcopodda avatar Apr 12 '21 11:04 marcopodda

  1. These two generation methods should be applicable in this article of yours?
  2. However, there are hundreds of thousands of Vocabulary on the model that does not use LFM. Will the generated model be problematic?

jkwang93 avatar Apr 12 '21 13:04 jkwang93

  1. More over, why choose BRICS instead of RECAP to break down molecules?Is it because RECAP cannot assemble fragments?

jkwang93 avatar Apr 13 '21 03:04 jkwang93