Dear Rdkit team,
My name is Velik Velikov and I am currently doing my bachelor's degree in
Bioinformatics.
My goal is to create a preference matrix from a set of molecules that
indicates how often one fragment reaction type is bound to another fragment
reaction type. Based on the preference values in the matrix, I want to
build a new molecule (retrosynthetically). To do this, I use BRICS to
obtain all fragments. I also tried using rdChemReaction and runReactants.
I am facing many problems in my implementation and after a week of reading
many threads in the forum, I decided to ask for help. So far this topic has
helped me a lot
https://sourceforge.net/p/rdkit/mailman/rdkit-discuss/thread/CAD4fdRSJc2F8NaPo1FBe%3DFoBaezjn2Nc5S26UnObazmK%3De0EpQ%40mail.gmail.com/#msg36610194
I tried to go from there and one of the problems I ran into here is that
when I get the fragments from rxn.RunReactants there is no information
about which reaction side was used. And some extra issues that didn't match
with my goal either.
So I am back to my initial idea. For example, after I gett all the
fragments from BRICSDecompose for the molecule Sorafenib
['[1*]C([1*])=O', '[1*]C([6*])=O', '[14*]c1cc([16*])ccn1',
'[16*]c1ccc(Cl)c([16*])c1', '[16*]c1ccc([16*])cc1', '[3*]O[3*]', '[5*]NC',
'[5*]N[5*]', '[8*]C(F)(F)F']
One can see every reaction type in every fragment. Let's say the first
fragment has reaction type 1 ('[1*]) twice. I would now like to know for
all the reaction types found, to which reaction type (in the other
fragment) there was a bond in the input molecule. Then I would update the
preference value for the two found reaction types in the matrix. And so on
for all 16 reaction types. I couldn't continue because I couldn't figure
out how to identify which fragment was connected to which fragment and on
which reaction side.
Would you please give me any tipps on how can I do that? Thanks in advance.
Best regards,
Velik Velikov
___
Rdkit-discuss mailing list
Rdkit-discuss@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/rdkit-discuss