This is a test version of Biostars. For the public version, visit https://www.biostars.org.
Best Practice For Extracting Gene Names From Pubmed Abstracts

I have a collection of several hundreds of Pubmed Abstracts. What I want to do is to extract all the gene names reported there.

What's the best way to go about it programmatically? Later I'd simply create a hash with pubmed ID as keys and gene names as its members.

pubmed
• 7,164 views • link
updated by Pierre Lindenbaum Pierre Lindenbaum • written by gundalav gundalav

6 answers

There is a webservice for whatizit: http://www.ebi.ac.uk/webservices/whatizit/helpws.jsp

it comes with an example:

String pipelineName = "whatizitSwissprot";
String pmid = "9879"; // The number is a Pubmed accession key
String xml = whatizit.queryPmid(pipelineName, pmid);
System.out.println(xml);
• 0 views • link
written by Pierre Lindenbaum Pierre Lindenbaum

You can get a list of NCBI Gene IDs based on Pubmed IDs using the GNAT webservice: http://textmining.ls.manchester.ac.uk:8081/. To call this webservice, provide a PMID list to the base URL with the pmid parameter as follows:

http://textmining.ls.manchester.ac.uk:8081/?pmid=21483786,21483692

For more information about this webservice, see this paper.

• 0 views • link
written by Casey Bergman Casey Bergman

Hi Thanks, Casey. Do you have any version that also recognize miRNA?

• 0 views • link
written by gundalav gundalav

At the time we made this version of the NCBI gene name dictionaries, mirna's were included in the Entrez gene. It looks like they are now though, so we should consider this in the next version of the GNAT dictionaries that are built.

• 0 views • link
written by Casey Bergman Casey Bergman

Reflect http://reflect.ws/ can be used programatically, and is designed to extract biological (and chemical) entities from text. It has a huge dictionary of synonyms sitting behind it, and could be useful for you - check out the "About" page REFLECT webpage for more info.

• 0 views • link
written by aidan-budd aidan-budd

There are several papers out there on pubmed text mining i.e. papers which report protein-protein interactions. DId you check the pubmed/google before posting?

Once you have a list of gene names, you can easily make a dictionary out of PMID and gene names in abstracts in python. Also have a look at Pubmed API.

• 0 views • link
written by Pappu Pappu

+1 to aidan-budd! +1 to the Reflect API developers for providing us with such a stable service

In the "About" page of Reflect you may find a link to the Reflect's REST API (http://reflect.ws/REST_API.html) "GetEntities" invoked with the "pmid" parameters provide you with a solution.

• 0 views • link
written by Evangelos • Evangelos

you should add this as a comment to the post

• 0 views • link
written by Istvan Albert Istvan Albert

(Added Feb 2014) http://www.ncbi.nlm.nih.gov/pubmed/23736528

Bioinformatics. 2013 Aug 1;29(15):1915-6. doi:

10.1093/bioinformatics/btt317. Epub 2013 Jun 4. BeCAS: biomedical concept recognition services and visualization. Nunes T1, Campos D,

Matos S, Oliveira JL.

it provides a nice REST-API: http://bioinformatics.ua.pt/becas

• 0 views • link
written by Pierre Lindenbaum Pierre Lindenbaum

Log in to answer this question.

More posts like this