Home >
Community >
What is the best scoring function for docking?
Upvote
26
Downvote
+ Docking
+ Molecular modeling
+ Qsar
+ Drug discovery
+ Cheminformatics and computational chemistry
+ Non-covalent interactions
+ Modeling and docking
+ Docking software
+ Cheminformatics
+ Intermolecular interactions
+ Computational biochemistry
+ Medical chemistry
+ Computational drug designing
Posted by
Mark Daly
What is the best scoring function for docking?
Consensus scoring works better than a randomly chosen single scoring function from a set. However usually the best single scoring function from the set outperforms the consensus; the problem is knowing which one might be best in advance.
Consensus scoring works better than a randomly chosen single scoring function from a set. However usually the best single scoring function from the set outperforms the consensus; the problem is knowing which one might be best in advance.
There is nothing like that becze each scoring function has it's own benefit and drawbacks based on parameters involve . So you can't determine which one you need to use before trying to asses it's validity with your docking or based on previous publications .
There is nothing like that becze each scoring function has it's own benefit and drawbacks based on parameters involve . So you can't determine which one you need to use before trying to asses it's validity with your docking or based on previous publications .
A general statement cannot be made. Even though, it can be said that scoring function with explicit solvation/desolvation penalties, charge tretment and buried hydrophobic paches terms; other than the normal interaction terms would outperforms in most instances. However, it becomes a trend to use score normalization strategy and the same has been asked in most standard journal. ArticleThe scoring bias in reverse docking and the score normalizat...
A general statement cannot be made. Even though, it can be said that scoring function with explicit solvation/desolvation penalties, charge tretment and buried hydrophobic paches terms; other than the normal interaction terms would outperforms in most instances. However, it becomes a trend to use score normalization strategy and the same has been asked in most standard journal. ArticleThe scoring bias in reverse docking and the score normalizat...
If you are using only one ligand against a particular protein then the best score should be selected on the basis of Glide emodel value which may or may not correspond to Glide score. But if you are ranking inhibitor or inhibitors against same protein then give preference to Glide score and not to glide emodel score for ranking
If you are using only one ligand against a particular protein then the best score should be selected on the basis of Glide emodel value which may or may not correspond to Glide score. But if you are ranking inhibitor or inhibitors against same protein then give preference to Glide score and not to glide emodel score for ranking
There are published benchmarks out there comparing different scoring functions. From what I heard, the concensus scoring approach based on multiple scoring functions seems to perform better than a single scoring function.
There are published benchmarks out there comparing different scoring functions. From what I heard, the concensus scoring approach based on multiple scoring functions seems to perform better than a single scoring function.
You should also remember that ranking compounds by their probability of being active (virtual screening) is, at the moment, a different problem to predicting the bound pose of a compound (pose prediction). So different scoring functions perform better for one or the other of these problems. Also scoring functions cannot predict binding affinity or binding free energy for two reasons: They calculate mostly enthalpic terms, and disregard entropy, particularly of the protein. Entropy is required, obviously, to estimate binding free energy. Scoring functions only know about the bound state of the protein-ligand system, not the unbound states of the protein and the ligand. BInding free energy can only be estimated using knowledge of the bound state and the unbound states of the binding partners.
You should also remember that ranking compounds by their probability of being active (virtual screening) is, at the moment, a different problem to predicting the bound pose of a compound (pose prediction). So different scoring functions perform better for one or the other of these problems. Also scoring functions cannot predict binding affinity or binding free energy for two reasons: They calculate mostly enthalpic terms, and disregard entropy, particularly of the protein. Entropy is required, obviously, to estimate binding free energy. Scoring functions only know about the bound state of the protein-ligand system, not the unbound states of the protein and the ligand. BInding free energy can only be estimated using knowledge of the bound state and the unbound states of the binding partners.
All of the answers provided so far are correct, there is no best scoring function for all possible cases. Different functions perform better in different cases, heres two possible cases a deeply buried hydrophobic pocket, versus a shallow groove site with lots of solvent accessibility a given function may do well in the first case and absolutely fail in the second. The best method is to test a variety on binding sites similar to what your interested in and if possible ligands similar to yours, at the bare minimum try some redocking runs using different scoring functions.
All of the answers provided so far are correct, there is no best scoring function for all possible cases. Different functions perform better in different cases, heres two possible cases a deeply buried hydrophobic pocket, versus a shallow groove site with lots of solvent accessibility a given function may do well in the first case and absolutely fail in the second. The best method is to test a variety on binding sites similar to what your interested in and if possible ligands similar to yours, at the bare minimum try some redocking runs using different scoring functions.
The best scoring function does not exist yet. What you can do is to make an external one yourself for the target you choose. For istance by making a 3-D QSAR model on docked structures or co-crystallized ligand.
The best scoring function does not exist yet. What you can do is to make an external one yourself for the target you choose. For istance by making a 3-D QSAR model on docked structures or co-crystallized ligand.
Consensus scoring works better than a randomly chosen single scoring function from a set. However usually the best single scoring function from the set outperforms the consensus; the problem is knowing which one might be best in advance.
Consensus scoring works better than a randomly chosen single scoring function from a set. However usually the best single scoring function from the set outperforms the consensus; the problem is knowing which one might be best in advance.
More
VOTE
There is nothing like that becze each scoring function has it's own benefit and drawbacks based on parameters involve . So you can't determine which one you need to use before trying to asses it's validity with your docking or based on previous publications .
There is nothing like that becze each scoring function has it's own benefit and drawbacks based on parameters involve . So you can't determine which one you need to use before trying to asses it's validity with your docking or based on previous publications .
More
VOTE
Docking and subsequent MD on docked complexes can give the suitability of scoring function used in docking program.
Docking and subsequent MD on docked complexes can give the suitability of scoring function used in docking program.
More
VOTE
A general statement cannot be made. Even though, it can be said that scoring function with explicit solvation/desolvation penalties, charge tretment and buried hydrophobic paches terms; other than the normal interaction terms would outperforms in most instances.
However, it becomes a trend to use score normalization strategy and the same has been asked in most standard journal.
Article The scoring bias in reverse docking and the score normalizat...
A general statement cannot be made. Even though, it can be said that scoring function with explicit solvation/desolvation penalties, charge tretment and buried hydrophobic paches terms; other than the normal interaction terms would outperforms in most instances.
However, it becomes a trend to use score normalization strategy and the same has been asked in most standard journal.
Article The scoring bias in reverse docking and the score normalizat...
More
VOTE
If you are using only one ligand against a particular protein then the best score should be selected on the basis of Glide emodel value which may or may not correspond to Glide score. But if you are ranking inhibitor or inhibitors against same protein then give preference to Glide score and not to glide emodel score for ranking
If you are using only one ligand against a particular protein then the best score should be selected on the basis of Glide emodel value which may or may not correspond to Glide score. But if you are ranking inhibitor or inhibitors against same protein then give preference to Glide score and not to glide emodel score for ranking
More
VOTE
There are published benchmarks out there comparing different scoring functions. From what I heard, the concensus scoring approach based on multiple scoring functions seems to perform better than a single scoring function.
There are published benchmarks out there comparing different scoring functions. From what I heard, the concensus scoring approach based on multiple scoring functions seems to perform better than a single scoring function.
More
VOTE
You should also remember that ranking compounds by their probability of being active (virtual screening) is, at the moment, a different problem to predicting the bound pose of a compound (pose prediction). So different scoring functions perform better for one or the other of these problems.
Also scoring functions cannot predict binding affinity or binding free energy for two reasons:
They calculate mostly enthalpic terms, and disregard entropy, particularly of the protein. Entropy is required, obviously, to estimate binding free energy.
Scoring functions only know about the bound state of the protein-ligand system, not the unbound states of the protein and the ligand. BInding free energy can only be estimated using knowledge of the bound state and the unbound states of the binding partners.
You should also remember that ranking compounds by their probability of being active (virtual screening) is, at the moment, a different problem to predicting the bound pose of a compound (pose prediction). So different scoring functions perform better for one or the other of these problems.
Also scoring functions cannot predict binding affinity or binding free energy for two reasons:
They calculate mostly enthalpic terms, and disregard entropy, particularly of the protein. Entropy is required, obviously, to estimate binding free energy.
Scoring functions only know about the bound state of the protein-ligand system, not the unbound states of the protein and the ligand. BInding free energy can only be estimated using knowledge of the bound state and the unbound states of the binding partners.
More
VOTE
All of the answers provided so far are correct, there is no best scoring function for all possible cases. Different functions perform better in different cases, heres two possible cases a deeply buried hydrophobic pocket, versus a shallow groove site with lots of solvent accessibility a given function may do well in the first case and absolutely fail in the second. The best method is to test a variety on binding sites similar to what your interested in and if possible ligands similar to yours, at the bare minimum try some redocking runs using different scoring functions.
All of the answers provided so far are correct, there is no best scoring function for all possible cases. Different functions perform better in different cases, heres two possible cases a deeply buried hydrophobic pocket, versus a shallow groove site with lots of solvent accessibility a given function may do well in the first case and absolutely fail in the second. The best method is to test a variety on binding sites similar to what your interested in and if possible ligands similar to yours, at the bare minimum try some redocking runs using different scoring functions.
More
VOTE
The best scoring function does not exist yet. What you can do is to make an external one yourself for the target you choose. For istance by making a 3-D QSAR model on docked structures or co-crystallized ligand.
The best scoring function does not exist yet. What you can do is to make an external one yourself for the target you choose. For istance by making a 3-D QSAR model on docked structures or co-crystallized ligand.
More
VOTE