Hi, thank you very much for sharing this impressive work.
I am currently working on a project that may involve CNS-related compounds, so your method is very interesting to me. Your clarification would be very helpful for me when thinking about dataset selection and applicability in my own study.
I may be misunderstanding part of the scoring design, but from my reading, PharmacoNet evaluates ligands at the pharmacophore/NCI level rather than the atom-pair level. My intuition is that this design may already alleviate some of the molecular-size bias that conventional atomistic scoring functions can show.
I am asking mainly to better understand the method, rather than to challenge the design. In particular, since larger molecules may contain more functional groups and potentially generate more matched pharmacophoric node pairs, I wanted to confirm whether this issue was considered during the scoring design.
Specifically, I would like to ask:
-
Was this issue considered during the design of the scoring function?
-
.Was the expectation that the current formulation would already mitigate this issue implicitly?
-
During validation, did you observe any correlation between the final score and molecular size-related descriptors such as heavy atom count, molecular weight, or number of pharmacophore features, especially among the top 1% ranked molecules?
My current understanding is that the pharmacophore-level abstraction may already reduce this effect compared with atomistic scoring, but I am not fully sure, so I wanted to ask for clarification.
Thank you again for your work. Your answer would be very helpful for my own study design, especially in database selection.
Hi, thank you very much for sharing this impressive work.
I am currently working on a project that may involve CNS-related compounds, so your method is very interesting to me. Your clarification would be very helpful for me when thinking about dataset selection and applicability in my own study.
I may be misunderstanding part of the scoring design, but from my reading, PharmacoNet evaluates ligands at the pharmacophore/NCI level rather than the atom-pair level. My intuition is that this design may already alleviate some of the molecular-size bias that conventional atomistic scoring functions can show.
I am asking mainly to better understand the method, rather than to challenge the design. In particular, since larger molecules may contain more functional groups and potentially generate more matched pharmacophoric node pairs, I wanted to confirm whether this issue was considered during the scoring design.
Specifically, I would like to ask:
Was this issue considered during the design of the scoring function?
.Was the expectation that the current formulation would already mitigate this issue implicitly?
During validation, did you observe any correlation between the final score and molecular size-related descriptors such as heavy atom count, molecular weight, or number of pharmacophore features, especially among the top 1% ranked molecules?
My current understanding is that the pharmacophore-level abstraction may already reduce this effect compared with atomistic scoring, but I am not fully sure, so I wanted to ask for clarification.
Thank you again for your work. Your answer would be very helpful for my own study design, especially in database selection.