I have downloaded 8000 PDB ids by using advanced search feature of PDB DB. I have used Ascorbic acid formula C6H8O6 to search for the PDB ids. Now I want to narrow down the list of PDB IDs. I have removed more that 500 PDB IDs which were mutated. Please suggest any other filter that I can use to reduce the no. of PDB IDs further?
Any help will be appreciated!
Thank you
That depends on what questions you wish to answer with your query. If you look for structures that actually contain ascorbate, you look in advanced search mode for Structure attribute: Comp Id (Nonpolymer Instance Feature Summary) equals "ASC" , https://www.rcsb.org/ligand/ASC , and you get 41 entries that actually contain ascorbate: 1E71, 1E72, 1E73, 1F9G, 1OAF, 1XID, 2X08, 2YLD, 2YLG, 3EKA, 3FXR, 3TGY, 3VLN, 3VXI, 3ZWI, 4CIP, 4GPC, 4GPH, 4JTP, 4O79, 4O7G, 4RP8, 4RP9, 4TVT, 4TWL
You can expand the selection to similar ligands, e.g. isoascorbic acid
If you, on the other hand, do a full-text search for "ascorbate" (759 hits) or "ascorbic" (1239 hits) you get all entries that mention the compound anywhere at all, including the abstract of the relevant publications. Only few of these structures do actually contain ascorbate.
If you enter "ascorbate" in the simple query window, you get a pulldown list of the contexts in which this keyword was found,
e.g. UniProt molecule name: e..g. Ascorbate peroxydase, in the structure name or additional structure keywords etc. Selecting the last item of this list, "in chemical synonyms", you get the 41 structures that actually contain ascorbate
If you do an advanced query for chemical: C6H8O6 you get 104 structures, comprising 8 different molecular entities with the same composition, which are mostly not related to ascorbate, e.g. TRICARBALLYLIC ACID
Thank you, Your answer helped me to cut down the redundancies. Is there any other search criteria which returns unique PDB IDs based on sequence similarity? I couldn't find it in the advance search.
Since many entries contain multiple chains, either because of multiple instances of the protein in the asymmetric unit, or because the entry contains a multiple chain complex, sequence similarity would not be uniquely defined.
If you search with a query sequence for sequence similarity, and request the return by "Polymer entities" rather than "Structures", you can request to get results grouped by xx% sequence identity, for xx=100%, 95%, 90%,70%, 50% or 30% sequence identity, displaying as groups or representatives
For this or any other search result, you can also download "Tabular results" item "Sequence Cluster", which gives you a CSV file containing the sequence cluster information for each chain in each structure of your result. You can open this file in Excel and further process as you see fit.
If you download the "sequence" item of the "Tabular results", it looks something like the attached screenshot.This is not really a format understood by sequence alignment programs. The easiest way is to convert it to a multi sequence fastA file, which is understood by sequence alignment programs. A FastA file contains for each sequence a header line starting with ">" and containing the sequence name and any additional information in a single line, followed by one or more lines containing the sequence information. I personally use a formula to gather the data I want to export from EXCEL into a single column using a formula:
=Left(A3,4)&"_"&left(C3,1)&"//"&G3
and copy the formula down the column.
I copy-paste the content of this column into an ASCII text editor (e.g. BBedit on a Mac or Notepad++ on Win) and replace the "//" by a linefeed character ( , \r , 0x0A (10 decimal) , representing a UNIX lie break).
The resulting text file is a valid FastA file, which I can further process with a sequence alignment program, e.g. Clustal omega http://www.clustal.org/omega/ or server https://www.ebi.ac.uk/Tools/msa/clustalo/
As you see in the example, I like to use the structure ID combined with the first chain label (if there are multiple redundant sequences in a pdb entry) as sequence identifier - this is my personal choice, you can use your own convention, as long as each sequence gets an unique identifier and the sequence name does not get too long, many alignment programs tend to truncate long names.
Thanks a lot Annemarie Honegger Ma'am!! I can now make fastA file for my proteins for sequence matching. I am using CD-HIT for clustering and comparing proteins based on sequence similarity.
- Badges
- Science topic
- Similar topics
- Phytochemistry
- Extracts