Notes
user@computer:~$represents an example terminal prompt name. Actual command/argument input is after the$.- Replace
[email protected]with your email address.\followed by>on the next line represents continued terminal input. You will need to delete the>symbol in order to run the scripts as a copy/paste into terminal.- You should validate your own EDirect scripts and results as there may be unintentional mistakes in these recipes. A convenient method is to compare your EDirect results to the NCBI Web interface search results: https://www.ncbi.nlm.nih.gov/.
We can use the EDirect function esearch to query PubMed. However, before trying to retrieve any of the results with efetch, it is a good idea to check that the count range is manageable (e.g., on the order of several thousand). In addition, see the EDirect Query Translation Instructions for how to use the -debug option to view how your query is interpreted in PubMed.
user@computer:~$ esearch -email [email protected] -db pubmed -query "hydrogel-based drug delivery"
<ENTREZ_DIRECT>
<Db>pubmed</Db>
<WebEnv>MCID...</WebEnv>
<QueryKey>1</QueryKey>
<Count>436</Count>
<Step>1</Step>
<Email>[email protected]</Email>
</ENTREZ_DIRECT>After deciding if the esearch query is appropriate, we can start to pipe the esearch results into other EDirect functions. For example, the below script first uses esearch to query PubMed for "hydrogel-based drug delivery", and then these results are piped (|) into efetch to retrieve the results as XML format. The efetch results are then piped to the xtract function where several bibliographic elements of the PubMed XML records are extracted into a table:
user@computer:~$ esearch -email [email protected] -db pubmed -query "hydrogel-based drug delivery" | \
> efetch -format xml | \
> xtract -pattern PubmedArticle -element MedlineCitation/PMID -first Author/LastName \
> Author/Initials ArticleTitle ISOAbbreviation PubDate/Year Volume Issue MedlinePgn
33424262 El-Masry SM Hydrogel-based matrices for controlled drug delivery of etamsylate: Prediction of in-vivo plasma profiles. Saudi Pharm J 2020 28 12 1704-1718
33398321 Chen W Magnetically actuated intelligent hydrogel-based child-parent microrobots for targeted drug delivery. J Mater Chem B 2021
33396629 Dehshahri A New Horizons in Hydrogels for Methotrexate Delivery. Gels 2020 7 1
33387892 Amiri M Hydrogel beads-based nanocomposites in novel drug delivery platforms: Recent trends and developments. Adv Colloid Interface Sci 2020 288 102316
33378390 Kloepping KC Triphenylphosphonium derivatives disrupt metabolism and inhibit melanoma growth in vivo when delivered via a thermosensitive hydrogel. PLoS One 2020 15 12 e0244540
33359482 Agarwal P Structural characterization and developability assessment of sustained release hydrogels for rapid implementation during preclinical studies. Eur J Pharm Sci 2021 158 105689
...
...tested on 2021.01.27, EDirect 14.4, total count was 436.
Note that if we want to extract out the DOIs, we can use the xtract -block option like this:
user@computer:~$ esearch -email [email protected] -db pubmed -query "hydrogel-based drug delivery" | \
> efetch -format xml | \
> xtract -pattern PubmedArticle -element MedlineCitation/PMID -first Author/LastName \
> Author/Initials ISOAbbreviation PubDate/Year Volume Issue MedlinePgn \
> -block ArticleId -if ArticleId@IdType -equals doi -doi ArticleId
33424262 El-Masry SM Saudi Pharm J 2020 28 12 1704-1718 https://doi.org/10.1016%2Fj.jsps.2020.10.016
33398321 Chen W J Mater Chem B 2021 https://doi.org/10.1039%2Fd0tb02384a
33396629 Dehshahri A Gels 2020 7 1 https://doi.org/10.3390%2Fgels7010002
33387892 Amiri M Adv Colloid Interface Sci 2020 288 102316 https://doi.org/10.1016%2Fj.cis.2020.102316
33378390 Kloepping KC PLoS One 2020 15 12 e0244540 https://doi.org/10.1371%2Fjournal.pone.0244540
33359482 Agarwal P Eur J Pharm Sci 2021 158 105689 https://doi.org/10.1016%2Fj.ejps.2020.105689
...
...tested on 2021.01.27, EDirect 14.4, total count was 436.
There is a lot going on with the last line of code that extracts out the DOIs: -block ArticleId -if ArticleId@IdType -equals doi -doi ArticleId. Let's look at part of a PubMed XML file to help interpret what is going on here:
...
...
<ArticleIdList>
<ArticleId IdType="pubmed">17630804</ArticleId>
<ArticleId IdType="doi">10.1021/jo071035l</ArticleId>
</ArticleIdList>
...
...The -block option limits the extraction to a particular section of the XML, in this case the ArticleId tags. The @ defines the desired IdType doi element attribute. Finally, the -doi is an xtract string option that prefixes https://doi.org/ before the extracted ArticleId doi. There is a more thorough explanation of -block and extracting out the DOIs with the -block option in the NLM Insider's Guide to Accessing NLM Data Part 4 and Entrez Programming Utilities Help Manual.
Similarly to the above script, we can specify particular fields to query within PubMed. The below script searches for "ionic liquids" in the MeSH term field ([MESH]) and "imidazolium" in all fields. Note that the internal quotes are escaped (\), which is sometimes necessary for the query to be interpreted correctly when using phrases.
user@computer:~$ esearch -email [email protected] -db pubmed -query "\"ionic liquids\"[MESH] AND imidazolium" | \
> efetch -format xml | \
> xtract -pattern PubmedArticle -element MedlineCitation/PMID -first Author/LastName \
> Author/Initials ISOAbbreviation PubDate/Year Volume Issue MedlinePgn \
> -block ArticleId -if ArticleId@IdType -equals doi -doi ArticleId
33396149 Hu LX Ecotoxicol Environ Saf 2021 208 111629 https://doi.org/10.1016%2Fj.ecoenv.2020.111629
33346267 Kaur M Phys Chem Chem Phys 2021 23 1 320-328 https://doi.org/10.1039%2Fd0cp04513f
33253998 Tashakkori P J Chromatogr A 2021 1635 461741 https://doi.org/10.1016%2Fj.chroma.2020.461741
33142384 Ren YM Zhonghua Lao Dong Wei Sheng Zhi Ye Bing Za Zhi 2020 38 10 767-769 https://doi.org/10.3760%2Fcma.j.cn121094-20191010-00483
33135708 Kumar S Phys Chem Chem Phys 2020 22 43 25255-25263 https://doi.org/10.1039%2Fd0cp04014b
32822985 Zuo L J Chromatogr A 2020 1628 461446 https://doi.org/10.1016%2Fj.chroma.2020.461446
32711338 Zunita M Bioresour Technol 2020 315 123864 https://doi.org/10.1016%2Fj.biortech.2020.123864
...
...tested on 2021.01.27, EDirect 14.4, total count was 1000.
The below script uses esearch to query PubMed for "Artificial Intelligence" in the [MESH] field and "drug discovery" in the [ALL] field. The records are then retrieved as XML format using the efetch function, followed by extracting out the journal names (IsoAbbreviation) using xtract. The xtract results are then piped to the EDirect alias function sort-uniq-count-rank, which sorts the data by highest frequency:
user@computer:~$ esearch -email [email protected] -db pubmed -query "\"Artificial Intelligence\"[MESH] AND \"drug discovery\"[ALL]" | \
> efetch -format xml | \
> xtract -pattern PubmedArticle -element ISOAbbreviation | \
> sort-uniq-count-rank
169 J Chem Inf Model
53 BMC Bioinformatics
49 PLoS One
40 Bioinformatics
39 Methods Mol Biol
33 Mol Pharm
32 Molecules
29 Sci Rep
28 Drug Discov Today
28 J Comput Aided Mol Des
24 J Med Chem
23 Expert Opin Drug Discov
23 Int J Mol Sci
19 Curr Top Med Chem
18 Mol Inform
17 Future Med Chem
16 Nucleic Acids Res
15 Nature
15 PLoS Comput Biol
14 IEEE/ACM Trans Comput Biol Bioinform
...
...tested on 2021.01.27, EDirect 14.4.
The below script uses esearch to query PubMed for ("university of alabama" AND tuscaloosa) in the affiliation field ([AFFL]). Tuscaloosa was added to limit the number of retrieved records associated with The University of Alabama at Birmingham and The University of Alabama at Huntsville. Another approach could have been to use the NOT operator: "(university of alabama[AFFL]) NOT (birmingham[AFFL] OR huntsville[AFFL])". However, the latter approach may eliminate any collaborative articles with these institutions (affiliation searches are challenging!). Next, the results were retrieved as XML using efetch, followed by piping these results to xtract to extract out the publication year (PubDate/Year) and sort by frequency with sort-uniq-count-rank. Note that a conditional statement was used in the xtract pattern to only extract results from articles if the affiliation contains both chemistry and tuscaloosa. The thought here was that this would limit the results (mostly) to author publications from The University of Alabama (Tuscaloosa) Department of Chemistry:
user@computer:~$ esearch -email [email protected] -db pubmed -query "(university of alabama[AFFL] AND tuscaloosa[AFFL])" | \
> efetch -format xml | \
> xtract -pattern PubmedArticle -if Affiliation -contains chemistry -and Affiliation -contains tuscaloosa -element PubDate/Year | \
> sort-uniq-count-rank
65 2015
64 2020
59 2017
53 2018
49 2019
45 2016
41 2014
35 2013
30 2012
28 2008
26 2006
23 2007
23 2010
20 2004
19 2003
18 2009
17 1999
17 2001
...
...tested on 2021.01.27, EDirect 14.4.
If instead we want to know individual Author numbers in PubMed instead of total publications by year, we can change the xtract pattern:
user@computer:~$ esearch -email [email protected] -db pubmed -query "(university of alabama[AFFL] AND tuscaloosa[AFFL])" | \
> efetch -format xml | \
> xtract -pattern Author -if Affiliation -contains chemistry -and Affiliation -contains tuscaloosa -element LastName Initials | \
> sort-uniq-count-rank
108 Dixon DA
53 Vasiliu M
34 Vincent JB
32 Rogers RD
22 Bowman MK
20 Fang Z
14 Grant DJ
14 Thanthiriwatte KS
13 Cassady CJ
12 Frantom PA
11 Chen M
11 Kelley SP
11 Shamshina JL
10 Kispert LD
10 Metzger RM
10 Papish ET
9 Gerlach DL
9 Li S
9 Timkovich R
8 Matus MH
...
...tested on 2021.01.27, EDirect 14.4.
Let's take a closer look at the conditional xtract pattern specifying to extract data only if the affiliation contains chemistry and tuscaloosa:
user@computer:~$ esearch -email [email protected] -db pubmed -query "(university of alabama[AFFL] AND tuscaloosa[AFFL])" | \
> efetch -format xml | \
> xtract -pattern Author -if Affiliation -contains chemistry -and Affiliation -contains tuscaloosa -element LastName Initials Affiliation
Rowe SJ Department of Chemistry and Biochemistry, University of Alabama, Tuscaloosa, Alabama, USA.
Mecaskey RJ Department of Chemistry and Biochemistry, University of Alabama, Tuscaloosa, Alabama, USA.
Nasef M Department of Chemistry and Biochemistry, University of Alabama, Tuscaloosa, Alabama, USA.
Talton RC Department of Chemistry and Biochemistry, University of Alabama, Tuscaloosa, Alabama, USA.
Sharkey RE Department of Chemistry and Biochemistry, University of Alabama, Tuscaloosa, Alabama, USA.
Halliday JC Department of Chemistry and Biochemistry, University of Alabama, Tuscaloosa, Alabama, USA.
...
...tested on 2021.01.27, EDirect 14.4.
With a quick look at the ~1000 results, it seemed like we extracted out the intended data, however, I did notice some false positive results. One example was an article from The University of Alabama (Tuscaloosa) Department of Biological Sciences with an external collaborator having "Chemistry" in the Institution name. Other errors could be what we unintentionally excluded such as any records that do not have Tuscaloosa in the affiliation field (i.e., only a partial address or zip code). These type of affiliation searches are tricky, so test often and think through the results carefully.
The elink function can retrieve associated cites and cited references for PubMed records. Cites are the available references in the article (i.e. bibliography list) and cited are references to the article. Not all PubMed articles have associated citation reference data. The available reference data are from the NIH Open Citation Collection Dataset.
To retrieve the number of cites for a PubMed article, we can use the elink function, followed by xtract to extract out the Count element:
user@computer:~$ esearch -email [email protected] -db pubmed -query "29978703[PMID]" | \
> elink -cites | \
> xtract -pattern ENTREZ_DIRECT -element Count
11tested on 2021.01.27, EDirect 14.4.
Add efetch to your script if you want to retrieve the records:
user@computer:~$ esearch -email [email protected] -db pubmed -query "29978703[PMID]" | \
> elink -cites | \
> efetch -format xml | \
> xtract -pattern PubmedArticle -element MedlineCitation/PMID -first Author/LastName \
> Author/Initials ISOAbbreviation PubDate/Year Volume Issue MedlinePgn \
> -block ArticleId -if ArticleId@IdType -equals doi -doi ArticleId
29382051 Stefanachi A Molecules 2018 23 2 https://doi.org/10.3390%2Fmolecules23020250
27709885 Stempel E Acc Chem Res 2016 49 11 2390-2402 https://doi.org/10.1021%2Facs.accounts.6b00265
26661053 James MJ Chemistry 2016 22 9 2856-81 https://doi.org/10.1002%2Fchem.201503835
26313158 Liu BY Org Lett 2015 17 17 4380-3 https://doi.org/10.1021%2Facs.orglett.5b02230
22969063 Han X Angew Chem Int Ed Engl 2012 51 41 10390-3 https://doi.org/10.1002%2Fanie.201205238
18620434 Martin R Acc Chem Res 2008 41 11 1461-73 https://doi.org/10.1021%2Far800036s
...
...tested on 2021.01.27, EDirect 14.4.
Getting the cited records only requires changing -cites to -cited:
user@computer:~$ esearch -email [email protected] -db pubmed -query "29978703[PMID]" | \
> elink -cited | \
> efetch -format xml | \
> xtract -pattern PubmedArticle -element MedlineCitation/PMID -first Author/LastName \
> Author/Initials ISOAbbreviation PubDate/Year Volume Issue MedlinePgn \
> -block ArticleId -if ArticleId@IdType -equals doi -doi ArticleId
32537619 Fernandes RA Chem Commun (Camb) 2020 56 61 8569-8590 https://doi.org/10.1039%2Fd0cc02659j
32317969 Lautié E Front Pharmacol 2020 11 397 https://doi.org/10.3389%2Ffphar.2020.00397
30707497 Ivanova OA Chem Rec 2019 https://doi.org/10.1002%2Ftcr.201800166
30259622 Tymann D Angew Chem Int Ed Engl 2018 57 47 15553-15557 https://doi.org/10.1002%2Fanie.201808578tested on 2021.01.27, EDirect 14.4.
We can answer some interesting questions with the NIH Open Citation Collection Data. For example, I noticed that the PubMed XML records for articles in J Cheminform contain a reference list for articles in PubMed. So, theoretically, if we query PubMed for J Cheminform, extract out all of the references, and sort these by frequency, we should get the most cited references in J Cheminform article bibliographies (caveat: in the available PubMed citation data).
In the below script, the xtract pattern creates a new line for each extracted reference citation PMID from the ArticleId with pubmed attribute field:
user@computer:~$ esearch -email [email protected] -db pubmed -query "J Cheminform[JOUR]" | \
> efetch -format xml | \
> xtract -pattern Reference -if ArticleId@IdType -equals pubmed -element ArticleId | \
> sort-uniq-count-rank | \
> head -n 20
90 20426451
83 21982300
80 21948594
65 12653513
47 11259830
40 10592235
38 16796559
38 27899562
37 26400175
36 8709122
33 17154509
33 19498078
30 16381955
29 21059682
29 23343401
28 15667143
28 24214965
27 17932057
27 21425294
27 22587354tested on 2021.01.27, EDirect 14.4.
Note that when quickly viewing all of the sorted results (~10,000 lines), I did see maybe a 100 or so entries with two PMIDs per line or a DOI and a PMID. Since we specifically defined the pubmed IdType attribute, it is not exactly clear to me yet why there would be some extra data in there. Perhaps it is a mistake or inconsistency in the J Cheminform PubMed XML records.
We can take a quick look at the top 10 cited references using a for loop::
user@computer:~$ for refs in \
> "20426451" \
> "21982300" \
> "21948594" \
> "12653513" \
> "11259830" \
> "10592235" \
> "16796559" \
> "27899562" \
> "26400175" \
> "8709122"
> do
> esearch -email [email protected] -db pubmed -query "$refs[PMID]" |
> efetch -format xml |
> xtract -pattern PubmedArticle -element MedlineCitation/PMID -first Author/LastName \
> Author/Initials ArticleTitle ISOAbbreviation PubDate/Year Volume Issue MedlinePgn \
> -block ArticleId -if ArticleId@IdType -equals doi -doi ArticleId
> sleep 1
> done
20426451 Rogers D Extended-connectivity fingerprints. J Chem Inf Model 2010 50 5 742-54 https://doi.org/10.1021%2Fci100050t
21982300 O'Boyle NM Open Babel: An open chemical toolbox. J Cheminform 2011 3 33 https://doi.org/10.1186%2F1758-2946-3-33
21948594 Gaulton A ChEMBL: a large-scale bioactivity database for drug discovery. Nucleic Acids Res 2012 40 Database issue D1100-7https://doi.org/10.1093%2Fnar%2Fgkr777
12653513 Steinbeck C The Chemistry Development Kit (CDK): an open-source Java library for Chemo- and Bioinformatics. J Chem Inf Comput Sci 43 2 493-500 https://doi.org/10.1021%2Fci025584y
11259830 Lipinski CA Experimental and computational approaches to estimate solubility and permeability in drug discovery and development settings. Adv Drug Deliv Rev 2001 46 1-3 3-26 https://doi.org/10.1016%2Fs0169-409x%2800%2900129-0
10592235 Berman HM The Protein Data Bank. Nucleic Acids Res 2000 28 1 235-42 https://doi.org/10.1093%2Fnar%2F28.1.235
16796559 Steinbeck C Recent developments of the chemistry development kit (CDK) - an open-source java library for chemo- and bioinformatics.Curr Pharm Des 2006 12 17 2111-20 https://doi.org/10.2174%2F138161206777585274
27899562 Gaulton A The ChEMBL database in 2017. Nucleic Acids Res 2017 45 D1 D945-D954 https://doi.org/10.1093%2Fnar%2Fgkw1074
26400175 Kim S PubChem Substance and Compound databases. Nucleic Acids Res 2016 44 D1 D1202-13 https://doi.org/10.1093%2Fnar%2Fgkv951
8709122 Bemis GW The properties of known drugs. 1. Molecular frameworks.J Med Chem 1996 39 15 2887-93 https://doi.org/10.1021%2Fjm9602928Another interesting question would be what is the most cited Journal in J Cheminform articles (in the available PubMed citation data)? In the below script, we take a similar approach to above, but instead of extracting out the PMIDs, we extract out the Citation element. The line cut -d "." -f 1 deletes any data after the Journal abbreviation (e.g., "Drug Discov Today. 2006 Dec;11(23-24):1046-53" becomes "Drug Discov Today").
user@computer:~$ esearch -email [email protected] -db pubmed -query "J Cheminform[JOUR]" | \
> efetch -format xml | \
> xtract -pattern Reference -element Citation | \
> cut -d "." -f 1 | \
> sort-uniq-count-rank
2274 J Chem Inf Model
1268 J Cheminform
1167 Nucleic Acids Res
930 J Med Chem
620 Bioinformatics
576 J Chem Inf Comput Sci
425 J Comput Aided Mol Des
381 BMC Bioinformatics
351 Drug Discov Today
252 Mol Inform
247 J Comput Chem
236 PLoS One
222 Nature
215 Nat Rev Drug Discov
202 Proc Natl Acad Sci U S A
181 Science
148 Anal Chem
147 Proteins
140 J Mol Graph Model
...
...tested on 2021.01.27, EDirect 14.4.
Note that there is some inconsistency in the citation formats results here as well that would need to be evaluated and cleaned up for a more thorough analysis. For example, some of the extracted Citations included author names and article titles, so the cut command deleting everything after the first . does not suffice for those data entries.
Here is an interesting script to retrieve the count of PubMed records by create date ([CRDT]) for each month of 2020. Since there are over 100,000 records added to PubMed every month, a strategy using efetch likely would not work (i.e., trying to retrieve 500,000+ records would take a long time).
user@computer:~$ for date in \
> "2020/01" \
> "2020/02" \
> "2020/03" \
> "2020/04" \
> "2020/05" \
> "2020/06"
> do
> esearch -email [email protected] -db pubmed -query "$date[CRDT]" |
> xtract -pattern ENTREZ_DIRECT -lbl "$date" -element Count
> sleep 1
> done
2020/01 108863
2020/02 107561
2020/03 106386
2020/04 124575
2020/05 121324
2020/06 124664tested on 2021.01.27, EDirect 14.4.
Let's say we wanted to know how many articles in J Chem Inf Model (indexed in PubMed) are available freely in PubMed Central. We can first get a count for J Chem Inf Model records in PubMed by querying PubMed in the Journal field ([JOUR]), followed by retrieving the records, extracting out the PubDate, and then sorting by frequency:
user@computer:~$ esearch -email [email protected] -db pubmed -query "J Chem Inf Model[JOUR]" | \
> efetch -format docsum | \
> xtract -pattern DocumentSummary -element PubDate | \
> cut -d " " -f 1 | \
> sort-uniq-count-rank | \
> sort -k2,2
216 2005
280 2006
246 2007
225 2008
268 2009
203 2010
297 2011
306 2012
300 2013
309 2014
247 2015
232 2016
283 2017
237 2018
490 2019
612 2020
65 2021tested on 2021.01.27, EDirect 14.4.
In the above script, the line cut -d " " -f 1 deletes any data appearing after the year and sort -k2,2 sorts the data by the second column. Next, we can add elink into our script to find the linked records in PubMed Central (pmc) from the Entrez link pubmed_pmc:
user@computer:~$ esearch -email [email protected] -db pubmed -query "J Chem Inf Model[JOUR]" | \
> elink -target pmc -name pubmed_pmc | \
> efetch -format docsum | \
> xtract -pattern DocumentSummary -element PubDate | \
> cut -d " " -f 1 | \
> sort-uniq-count-rank | \
> sort -k2,2
1 2005
3 2006
7 2007
14 2008
31 2009
26 2010
62 2011
38 2012
55 2013
59 2014
38 2015
32 2016
39 2017
43 2018
57 2019
51 2020tested on 2021.01.27, EDirect 14.4.
Note that if you have a query returning tens of thousands of results, you would likely want to use a strategy without efetch, such as adding a date into your esearch query, followed by extracting out the count element from the XML.