Datasets:
license: cc-by-4.0
language:
- en
size_categories:
- 10K<n<100K
task_categories:
- feature-extraction
tags:
- Structural biology
- Bioinformatics
- Machine learning
- Conformation
- Conformational state
- Monomeric
- Training data
- Benchmark
- Manually curated
pretty_name: Curated dataset of protein chain conformational states
Schema description:
The manually curated dataset of open-closed monomers is included here as benchmarking_monomeric_open_closed_conformers.csv
.
Column descriptions:
Schema description:
The manually curated dataset of open-closed monomers is included here as benchmarking_monomeric_open_closed_conformers.csv
.
Column descriptions:
UNP_ACC
| UniProt accession codeUNP_START
| Start of UniProt sequence for given PDBe entriesUNP_END
| End of UniProt sequence for given PDBe entriesPDBe_ID
| Protein Data Bank codeCHAIN_ID
| Author declared chain ID (char
)label_asym_id
| Programmatically assigned chain ID (char
)CONFORMER_ID
| Unique code for PDBe entries with distinct conformation, corresponding to a given UniProt accessionCONFORMER_DESCR
| Short description of conformation, based on depositor's assessment of the protein/conformationLIT_CONFIRMED
| True/false value based on whether a publication (scientific literature) was available for manually curating clusters. NB: Clusters with 0 in this field should be used with caution.ALT_CONFORMER_ID
| Where the publication for a structure is currently outstanding, an executive decision on the conformation classification is made. Where the literature is not explicit on the features of a given conformation, the second most suitableCONFORMER_ID
is provided in this column. Blank cells have no other likely conformation assignmnt and are therefore the same as inCONFORMER_ID
.ALT_CONFORMER_DESCR
| Description for conformation in alternative conformation ID.
Curation process
As of 09 Mar 2022, a manually curated dataset of monomeric protein conformations has been collated, containing 'open'-'closed' pairs as well as intermediary states defined by the authors of the entry.
- The PDBe was queried, through its Oracle DB, to find PDBe entries with 100 % sequence identity for a UniProt segment in both 'open' and 'closed' conformations, as stated in the entry's
TITLE
field. The query used:
select b.accession, b.unp_start, b.unp_end, a.id, a.title, d.id, d.title
from entry a, unp_entity b, unp_entity c, entry d, pdb_assembly e
where a.title like ‘%open%’ and d.title like ‘%close%’
and a.id = b.entry_id and d.id = c.entry_id and a.id != d.id
and b.accession = c.accession
and b.unp_start = c.unp_start
and b.unp_end = c.unp_end
and a.id = e.entry_id
and e.type = ‘homo’
and e.name = ‘monomer’
was written by Dr Sameer Velankar.
These results were cleaned to remove entries with 'open' or 'close' substrings in their
TITLE
fields that did not refer to conformation. The 'open' substring often appeared in ligand names in the entries'TITLE
field, such as in dichlorido(1,3-dimethylbenzimidaz ol-2-ylidene)(eta5-pentamethylcyclopentadienyl)rhodium(III) and 'close' in terms like discloses.All remaining entries were then manually curated by investigating each PDBe entry's corresponding publication, where available.
Additional PDBe entries submitted by the authors, which were missed in the original search due to a lack of 'open' or 'close' substrings in their
TITLE
field but stated as fitting one of the states in the publication, were added.For some UniProt accessions, intermediary conformations were reported by the authors and these were noted in the dataset under the
CONFORMER_DESCR
column.Entries deposited in monomeric form but solved as a multimeric complex were also removed.
PDBe entries, now clustered by author-stated conformation, were cross-referenced against the PDBe-KB's existing clustering algorithm (available on the Aggregate Views of Proteins page) to assess current conformer clustering success. These results are currently awaiting publication.