I want to choose a good PDB file for docking my compounds intro a MCF-7 and MDA-MB231 cell line receptors.
Remember that a crystal structure is a model, built to explain experimental data, which is electron density. The question you want to ask is what is the best model available.
The best answer to the quality of a structure question is contained in this paper
Warren et al Drug Discovery Today 2012
in short
Electron density must be available, a model deposited in the PDB without electron density cannot be checked for quality.
Resolution is NOT a quality criterion, resolution tells you about how much data was gathered in the experiment, not how well the model fits the data.
The model quality can be assessed using Rfree and R, the lower the better.
The B factors of atoms in a model are usually NOT experimental observables; at the resolution of most structures in the PDB (> 2.3A) there is not enough data to assign a B factor, so the B factor is a free parameter in the refinement, not a property of atoms in the model. Also, B factors CANNOT be compared between structures that are solved using different software or protocols, so comparing B factors as a quality criterion for a model simply does not work.
Dear,
Following are few points to keep in mind while selecting a protein structure from PDB
1. Resolution must be minimum possible ( This will ensure the better quality of protein structure.)
2. Domain completeness. Examine your PDB structure and confirm the under study domain full structure availability. Partial domain will lead to false interpretations.
3. Variant /Mutations. According to your case study look for whether your structure is a wild type or a mutant/variant. In case of a mutant structure requirement you may have to introduce required mutations manually and model them.
4. Side Chain Completeness ( is of secondary importance). Structures determined through old techniques might have (Not always) missing side chains due to flaw in tech or manual error. Right 3D confirmation of Side chains is critical in small ligand binding thus ensure their completeness. As a possible solution you may look for latest structure availability of the same.
5. Ligand /Crystalline Water / Co factor presence. To get out the right docking result removal (As per case study ) of these elements form PDB file is important.
Wishes.
Also browse https://www.researchgate.net/post/For_docking_what_are_the_criteria_to_choose_the_best_protein_structure_from_PDB_to_be_used_as_the_target
for detailed discussion.
A while ago I download all the pdb with low resolution and exported to a database, them I collected the scores. The result is a quickly way to list the pdb based on the scores.
https://www.a00s.com/a00s_20141223/
user: [email protected]
pass: bm
just be aware that this pdb are not really reliable, some times even with good score have some weird spaces, some times protonate at ph 7, other times with the publication that doesnt match with the pdb file.
Taking in to account the opinion of Rajnikant Namdeo and Thiago Benazzi Maia, I would like to add about the B factor in the crystallographic data, you need to choose a PDB with the B factor lower in your specific region of protein for the study (active site, for example), because the B factor is related with the motion (real position) of the residues of the protein in that area. If you choose a PDB with a high B factor, you will obtain a bad prediction of the binding mode.
Remember that a crystal structure is a model, built to explain experimental data, which is electron density. The question you want to ask is what is the best model available.
The best answer to the quality of a structure question is contained in this paper
Warren et al Drug Discovery Today 2012
in short
Electron density must be available, a model deposited in the PDB without electron density cannot be checked for quality.
Resolution is NOT a quality criterion, resolution tells you about how much data was gathered in the experiment, not how well the model fits the data.
The model quality can be assessed using Rfree and R, the lower the better.
The B factors of atoms in a model are usually NOT experimental observables; at the resolution of most structures in the PDB (> 2.3A) there is not enough data to assign a B factor, so the B factor is a free parameter in the refinement, not a property of atoms in the model. Also, B factors CANNOT be compared between structures that are solved using different software or protocols, so comparing B factors as a quality criterion for a model simply does not work.
I agree with Rajnikant and Paul's comments. If you are able to get a PDB file with an inhibitor bound, that will help you to get a good idea on how your molecules will bind. If the solved structure has an inhibitor with similar chemical features as your compounds, that's even better.
- Badges
- Science topic