Stereo models explained

Posted by
Róbert Wagner
on 2013-09-03

Stereo models explained

Stereo models determine the way stereo chemistry is calculated, which affects stereo matching. The differences centre around the issue of symmetry. Our documentation has improved with additional examples and better descriptions yet it seems users can sometimes still be confused. This time we try to illustrate stereo models through three likely conversations we would have when explaining to users. Hope this helps with understanding.

Cast:

  • Poor ChemAxon Developer (PCD)
  • Chemist liking global stereo model (C♥GLOB)
  • Chemist liking comprehensive stereo model (C♥COMP)
  • Chemist liking local stereo model (C♥LOC)

 

Conversation with Global Chemist

 

PCD: Hi, how can I help you?

C♥GLOB: I would like a stereo matching behavior that doesn't consider symmetric structures as stereo centers, so:

 

  1. Tiny label shouldn’t match Tiny label 5 and Tiny label 3
  2. but Tiny label 9 should be handled as tiny label 11, and match tiny label 11, Tiny label 4 and itself as well.

PCD: I get the point of (a), but regarding (b): Tiny label 9 has a wedge while tiny label 11 hasn’t, so why should they match?

C♥GLOB: I don’t care about Tiny label 9's wedge, it is the same as Tiny label 4, because of symmetry. Just rotate the structure. Hence tiny label 11, Tiny label 9, and Tiny label 4 all mean the same molecule.

PCD: Now I got what you want, thanks!

Conversation with Comprehensive Chemist

C♥COMP: Hi, I would like a search that considers symmetric wedge/hatch signed stereo centers on the query side.

PCD: Why? tiny label 11, Tiny label 9, and Tiny label 4 are the same molecules, as C♥GLOB said!

C♥COMP: As full molecules yes, but I talk about the query side, and in case of substructure searching. A symmetric structure can match on an asymmetric one like tiny label 11 on Tiny label 10. By drawing query Tiny label 9 differently from tiny label 11, I tell that I’m interested in chiral centers so Tiny label 10 is OK, but Tiny label 2 is not.

PCD: Oh, I see. But wait, the last chiral target will be matched by Tiny label 4 as well!

C♥COMP: Yes, but the query should flip, which should be reflected in the hit atom mapping.

PCD: Of course, we can do that! Thank you!

C♥COMP: Hey, don't be hasty! Why don't we consider target symmetry? Tiny label 7 and Tiny label 8 mean the same chemical structure! I want the first to be hit as well!

PCD: You just wanted to distinguish stereo marked and flat query atoms!

C♥COMP: On the query side yes, because it can be a substructure, but the target is a whole molecule! Why do you require the user to put a wedge on a symmetric target structure, if it means the same as a flat one ( tiny label 11 and Tiny label 9 )?

PCD: Oh, I see, so symmetric target structures should always be hit by wedge/hatch marked or flat query structures?

C♥COMP: Yes! It’s incredible that you’ve understood it so slowly!

PCD: Sorry, we will fulfill your request.

Conversation with Local Chemist

C♥LOC: I would like to match stereo marked queries on stereo marked targets.

PCD: Yeah, that’s fine, but what about symmetric queries?

C♥LOC: I don’t care about symmetry. Consider the given stereo marking always.

PCD: On the query side I understand it based on C♥COMP’s points, but don’t tiny label 11 and Tiny label 9 mean the same targets?

C♥LOC: No, I only care about marking when searching with Tiny label 9. In this case if the target is not marked, it should be thrown out. For example, I would like to distinguish queries from each other. Like in a query collection (author's remark: see default stereo model for query tables), do you think tiny label 11 and Tiny label 9 are the same queries?

PCD: No, by no means. I see your point, of course we can do this.

C♥LOC: Thanks, bye.

Conclusion

 

As you could see chemists vary regarding their requests about the desired behavior. Furthermore in reality the same user may want to have different models depending on:

  • searching between full structures (duplicate search, global model)
  • substructure search (comprehensive model)
  • comparing query structures (duplicate search on query tables)

which is exactly what we provide in our default stereo model selection. Whatever model we choose in the automatic method, you can override the used stereo model in search options.

 

Stereo models determine the way stereo chemistry is calculated, which affects stereo matching. The differences centre around the issue of symmetry. Our documentation has improved with additional examples and better descriptions yet it seems users can sometimes still be confused. This time we try to illustrate stereo models through three likely conversations we would have when explaining to users. Hope this helps with understanding.

Cast:

  • Poor ChemAxon Developer (PCD)
  • Chemist liking global stereo model (C♥GLOB)
  • Chemist liking comprehensive stereo model (C♥COMP)
  • Chemist liking local stereo model (C♥LOC)

 

Conversation with Global Chemist

 

PCD: Hi, how can I help you?

C♥GLOB: I would like a stereo matching behavior that doesn't consider symmetric structures as stereo centers, so:

 

  1. Tiny label shouldn’t match Tiny label 5 and Tiny label 3
  2. but Tiny label 9 should be handled as tiny label 11, and match tiny label 11, Tiny label 4 and itself as well.

PCD: I get the point of (a), but regarding (b): Tiny label 9 has a wedge while tiny label 11 hasn’t, so why should they match?

C♥GLOB: I don’t care about Tiny label 9's wedge, it is the same as Tiny label 4, because of symmetry. Just rotate the structure. Hence tiny label 11, Tiny label 9, and Tiny label 4 all mean the same molecule.

PCD: Now I got what you want, thanks!

Conversation with Comprehensive Chemist

C♥COMP: Hi, I would like a search that considers symmetric wedge/hatch signed stereo centers on the query side.

PCD: Why? tiny label 11, Tiny label 9, and Tiny label 4 are the same molecules, as C♥GLOB said!

C♥COMP: As full molecules yes, but I talk about the query side, and in case of substructure searching. A symmetric structure can match on an asymmetric one like tiny label 11 on Tiny label 10. By drawing query Tiny label 9 differently from tiny label 11, I tell that I’m interested in chiral centers so Tiny label 10 is OK, but Tiny label 2 is not.

PCD: Oh, I see. But wait, the last chiral target will be matched by Tiny label 4 as well!

C♥COMP: Yes, but the query should flip, which should be reflected in the hit atom mapping.

PCD: Of course, we can do that! Thank you!

C♥COMP: Hey, don't be hasty! Why don't we consider target symmetry? Tiny label 7 and Tiny label 8 mean the same chemical structure! I want the first to be hit as well!

PCD: You just wanted to distinguish stereo marked and flat query atoms!

C♥COMP: On the query side yes, because it can be a substructure, but the target is a whole molecule! Why do you require the user to put a wedge on a symmetric target structure, if it means the same as a flat one ( tiny label 11 and Tiny label 9 )?

PCD: Oh, I see, so symmetric target structures should always be hit by wedge/hatch marked or flat query structures?

C♥COMP: Yes! It’s incredible that you’ve understood it so slowly!

PCD: Sorry, we will fulfill your request.

Conversation with Local Chemist

C♥LOC: I would like to match stereo marked queries on stereo marked targets.

PCD: Yeah, that’s fine, but what about symmetric queries?

C♥LOC: I don’t care about symmetry. Consider the given stereo marking always.

PCD: On the query side I understand it based on C♥COMP’s points, but don’t tiny label 11 and Tiny label 9 mean the same targets?

C♥LOC: No, I only care about marking when searching with Tiny label 9. In this case if the target is not marked, it should be thrown out. For example, I would like to distinguish queries from each other. Like in a query collection (author's remark: see default stereo model for query tables), do you think tiny label 11 and Tiny label 9 are the same queries?

PCD: No, by no means. I see your point, of course we can do this.

C♥LOC: Thanks, bye.

Conclusion

 

As you could see chemists vary regarding their requests about the desired behavior. Furthermore in reality the same user may want to have different models depending on:

  • searching between full structures (duplicate search, global model)
  • substructure search (comprehensive model)
  • comparing query structures (duplicate search on query tables)

which is exactly what we provide in our default stereo model selection. Whatever model we choose in the automatic method, you can override the used stereo model in search options.