I'm interested in retrieving the list of PubMed article identifiers (PMIDs) of all the articles in PubMed that is associated with a MeSH term like "Breast Neoplasms". I wish to repeat this for a number of MeSH terms, which therefore makes direct web-queries at http://www.ncbi.nlm.nih.gov/pubmed/ too painful.
Is there a way NCBI's E-utilities to do this efficiently from the linux command-line?
•
written
by
Arjun Krishnan •
Used ncbi esearch with the MESH field modifier (as defined in http://eutils.ncbi.nlm.nih.gov/entrez/eutils/einfo.fcgi?db=pubmed)
$ curl -s 'http://eutils.ncbi.nlm.nih.gov/entrez/eutils/esearch.fcgi?db=pubmed&term=%22Breast+Neoplasms%22%5BMESH%5D' | xmllint --format -
<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE eSearchResult PUBLIC "-//NLM//DTD esearch 20060628//EN" "http://eutils.ncbi.nlm.nih.gov/eutils/dtd/20060628/esearch.dtd">
<eSearchResult>
<Count>221299</Count>
<RetMax>20</RetMax>
<RetStart>0</RetStart>
<IdList>
<Id>25668825</Id>
<Id>25668824</Id>
<Id>25668823</Id>
<Id>25668822</Id>
<Id>25647216</Id>
<Id>25647215</Id>
<Id>25647190</Id>
<Id>25603628</Id>
<Id>25597209</Id>
<Id>25596051</Id>
<Id>25596048</Id>
<Id>25585789</Id>
<Id>25585788</Id>
<Id>25585780</Id>
<Id>25585779</Id>
<Id>25585778</Id>
<Id>25585328</Id>
<Id>25585323</Id>
<Id>25577824</Id>
<Id>25568923</Id>
</IdList>
<TranslationSet/>
<TranslationStack>
<TermSet>
<Term>"Breast Neoplasms"[MESH]</Term>
<Field>MESH</Field>
<Count>221299</Count>
<Explode>Y</Explode>
</TermSet>
<OP>GROUP</OP>
</TranslationStack>
<QueryTranslation>"Breast Neoplasms"[MESH]</QueryTranslation>
</eSearchResult>
Entrez Direct is another option
esearch -db pubmed -query "Burkholderia pseudomallei/metabolism[MESH]" | efetch -format uid
24866793
24626296
24595140
24502667
24462575
...
curl -vs 'http://eutils.ncbi.nlm.nih.gov/entrez/eutils/esearch.fcgi?db=pubmed&term=breast+neoplasms&retmode=text&retmax=1000' 2>&1 | grep "^<Id>"
should give you just the IDs with HTML tags. Use sed to remove the tags and transform in any way you'd like to see it.
Please note: I've set retmax to the 1000. You can change it to get IDs in batches of your preferred size. Update from Pierre is that this number cannot be >100,000.
Also, curl -vs writes to STDOUT without progress bar, 2>&1 combines STDERR and STDOUT so you can process the output directly through pipes.
FYI max(retmax)= 100,000
•
written
by
Pierre Lindenbaum
Log in to answer this question.
