TREC 2007 Genomics Track Data and Tools

Last updated - April 6, 2015

This page lists the files that are in the data distribution and the evaluation tools for the TREC 2007 Genomics Track. Additional detail about these files can be found in the 2007 track protocol. The data files themselves are available in the 2006 data portion of the Web site.

This page provides information about the following categories of data and tools for the TREC 2007 Genomics track:

Documents
Topics
Relevance Judgments
Evaluation Tools

Documents

The document collection used in 2007 is the same as that used for the TREC 2006 Genomics Track. It consists of full-text HTML documents from 49 journals who publish electronically via Highwire Press that granted permission for research use of their articles in the Genomics Track. The documents have been obtained by a Web crawl of the Highwire site, with postprocessing to eliminate as much non-article material as we could. Please note that we promised the publishers that this material would only be used for research purposes and would not be posted on public Web sites.

The full collection contains 162,259 documents from the 49 journals. The files are available on the protected portion of the track Web site in WinZip format. There are 59 .zip files, one for each journal, with the exception of the large Journal of Biological Chemistry that has its data in 10 files (one for each year). The 59 .zip files total about 3 GB in size. The full collection is about 12.3 GB when uncompressed.

There is also a text file, metadata.txt (Windows ASCII format, 11.9 MB), which lists the original URL of the article, the file name in our collection, and the file size in kilobytes is also available. Note that the name of each document file is its PMID plus the extension .html, which facilitates accessing the associated MEDLINE record as described below. Here is a sample of the text file:

http://www.rnajournal.org/cgi/content/full/9/1/1     12554869.html    73.4
http://www.rnajournal.org/cgi/content/full/9/1/9     12554870.html    46.43
http://www.rnajournal.org/cgi/content/full/9/1/14    12554871.html    38.17

The table below lists the .zip file name, size in MB, number of documents, and journal URL for each file in our collection.

Journal Name	File Name	File Size (MB)	Number of Docs	Journal URL
American Journal of Epidemiology	ajepidem.zip	24	1777	aje.oxfordjournals.org
American Journal of Physiology - Cell Physiology	ajpcell.zip	62	2906	ajpcell.physiology.org
American Journal of Physiology - Endocrinology And Metabolism	ajpendometa.zip	48	2462	ajpendo.physiology.org
American Journal of Physiology - Gastrointestinal and Liver Physiology	ajpgastro.zip	48	2472	ajpgi.physiology.org
American Journal of Physiology - Heart and Circulatory Physiology	ajpheart.zip	99	5170	ajpheart.physiology.org
American Journal of Physiology - Lung Cellular and Molecular Physiology	ajplung.zip	48	2426	ajplung.physiology.org
American Journal of Physiology - Renal Physiology	ajprenal.zip	39	1897	ajprenal.physiology.org
Alcohol and Alcoholism	alcohol.zip	9.7	657	alcalc.oxfordjournals.org
Journal of Andrology	andrology.zip	7.1	482	www.andrologyjournal.org
Annals of Oncology	annonc.zip	16	1273	annonc.oxfordjournals.org
British Journal of Anaesthesia	bjanast.zip	21	1843	bja.oxfordjournals.org
The British Journal of Psychiatry	bjp.zip	17	1531	bjp.rcpsych.org
Blood	blood.zip	209	11291	www.bloodjournal.org
Carcinogenesis	carcinogenesis.zip	36	2022	carcin.oxfordjournals.org
Cerebral Cortex	cercor.zip	22	917	cercor.oxfordjournals.org
Development	development.zip	62	2402	dev.biologists.org
Diabetes	diabetes.zip	37	2156	diabetes.diabetesjournals.org
Endocrinology	endocrinology.zip	104	5517	endo.endojournals.org
European Heart Journal	euroheartj.zip	15	1160	eurheartj.oxfordjournals.org
Glycobiology	glycobiology.zip	15	719	glycob.oxfordjournals.org
Human Reproduction	humanrep.zip	50	3784	humrep.oxfordjournals.org
Human Molecular Genetics	humolgen.zip	58	3105	hmg.oxfordjournals.org
International Journal of Epidemiology	ijepidem.zip	13	1203	ije.oxfordjournals.org
International Immunology	intimm.zip	23	1175	intimm.oxfordjournals.org
Journal of Antimicrobial Chemotherapy	jantichemo.zip	29	2720	jac.oxfordjournals.org
Journal of Applied Physiology	jappliedphysio.zip	105	5751	jap.physiology.org
Journal of Biological Chemistry	jbc-1995.zip	74	4368	www.jbc.org
Journal of Biological Chemistry	jbc-1996.zip	33	4733	www.jbc.org
Journal of Biological Chemistry	jbc-1997.zip	60	3098	www.jbc.org
Journal of Biological Chemistry	jbc-1998.zip	59	2918	www.jbc.org
Journal of Biological Chemistry	jbc-1999.zip	49	2432	www.jbc.org
Journal of Biological Chemistry	jbc-2000.zip	111	5361	www.jbc.org
Journal of Biological Chemistry	jbc-2001.zip	69	3262	www.jbc.org
Journal of Biological Chemistry	jbc-2002.zip	119	5539	www.jbc.org
Journal of Biological Chemistry	jbc-2003.zip	76	3510	www.jbc.org
Journal of Biological Chemistry	jbc-2004.zip	132	6214	www.jbc.org
Journal of Biological Chemistry	jbc-2005.zip	109	4886	www.jbc.org
The Journal of Cell Biology	jcb.zip	93	3996	www.jcb.org
Journal of Clinical Endocrinology & Metabolism	jclinicalendometa.zip	6.9	758	jcem.endojournals.org
Journal of Cell Science	jcs.zip	54	2417	jcs.biologists.org
Journal of Experimental Biology	jexpbio.zip	41	1911	jeb.biologists.org
Journal of Experimental Medicine	jexpmed.zip	70	3492	www.jem.org
The Journal of General Physiology	jgenphysio.zip	25	1014	www.jgp.org
Journal of General Virology	jgenviro.zip	40	2375	vir.sgmjournals.org
Journal of Histochemistry and Cytochemistry	jhistocyto.zip	24	1592	www.jhc.org
Journal of the National Cancer Institute	jnci.zip	34	3214	jncicancerspectrum.oxfordjournals.org
Journal of Neurophysiology	jneuro.zip	68	2874	jn.physiology.org
Molecular & Cellular Proteomics	mcp.zip	9.5	426	www.mcponline.org
Microbiology	microbio.zip	46	2400	mic.sgmjournals.org
Molecular Biology and Evolution	molbiolevol.zip	25	1303	mbe.oxfordjournals.org
Molecular Endocrinology	molendo.zip	36	1610	mend.endojournals.org
Molecular Human Reproduction	molhumanrep.zip	14	817	molehr.oxfordjournals.org
Nucleic Acids Research	nar.zip	126	7606	nar.oxfordjournals.org
Nephrology Dialysis Transplantation	nephrodiatransp.zip	38	3629	ndt.oxfordjournals.org
Protein Engineering Design and Selection	peds.zip	15	834	peds.oxfordjournals.org
Physiological Genomics	physiogenomics.zip	13	656	physiolgenomics.physiology.org
Rheumatology	rheumatolgy.zip	21	1985	rheumatology.oxfordjournals.org
RNA	rna.zip	11	544	www.rnajournal.org
Toxicological Sciences	toxsci.zip	33	1667	toxsci.oxfordjournals.org

We have discovered some issues with the full-text collection. One of these was fixed: the original files for the journal Blood were found to be lacking titles and author names and affiliations. We have corrected this and uploaded a new .zip file (time stamped 5-20-06, renaming the old one to blood-old.zip). We have also found the other following issues, for which we do not plan fixes:

There are some HTML files that are empty, i.e., contain no text. These do have corresponding MEDLINE records. However, they should not generate retrieved passages, since there are no passages in them. Therefore they should not influence the results.
There are some journals where a specific issue or year of a journal is in the collection but some articles are missing. This is most likely an artifact of the Web-crawling process we used, and errors either within our code or in Highwire documents. In any case, we consider the collection to be fixed and closed, so if an article is not in the collection, it should not be retrieved.

In addition to the full-text data, the NLM has been kind enough to provide us with both ASCII and XML format collections of all the MEDLINE references for the full-text documents in our Highwire collection.

There are a couple caveats about this data. First, we have identified 1,767 instances (about 1% of the 162K documents) where the Highwire file PMID is invalid. We have investigated the problem and found that for all of instances we checked, the problem was the original Highwire file having an incorrect PMID. In other words, the error is in the Highwire data, not a result of our processing. For this reason, we have decided not to delete these files from the collection. They represent, in our view, normal dirty data. Second, there must also likely be incorrect PMIDs in the Highwire data that happen to map to valid PMIDs. Below is a table of the number of erroroneous PMIDs we identified per journal.

Journal	PMID Errors
American Journal of Epidemiology	85
American Journal of Physiology - Cell Physiology	5
American Journal of Physiology - Endocrinology And Metabolism	3
American Journal of Physiology - Gastrointestinal and Liver Physiology	2
American Journal of Physiology - Heart and Circulatory Physiology	13
American Journal of Physiology - Lung Cellular and Molecular Physiology	7
American Journal of Physiology - Renal Physiology	7
Alcohol and Alcoholism	22
Journal of Andrology	26
Annals of Oncology	33
British Journal of Anaesthesia	13
The British Journal of Psychiatry	53
Blood	53
Carcinogenesis	25
Cerebral Cortex	10
Development	36
Diabetes	2
Endocrinology	35
European Heart Journal	5
Glycobiology	1
Human Reproduction	86
Human Molecular Genetics	15
International Journal of Epidemiology	83
International Immunology	2
Journal of Antimicrobial Chemotherapy	21
Journal of Applied Physiology	231
Journal of Biological Chemistry	74
The Journal of Cell Biology	48
Journal of Clinical Endocrinology & Metabolism	68
Journal of Cell Science	40
Journal of Experimental Biology	30
Journal of Experimental Medicine	24
The Journal of General Physiology	14
Journal of General Virology	2
Journal of Histochemistry and Cytochemistry	1
Journal of the National Cancer Institute	238
Journal of Neurophysiology	0
Molecular & Cellular Proteomics	10
Microbiology	10
Molecular Biology and Evolution	12
Molecular Endocrinology	49
Molecular Human Reproduction	17
Nucleic Acids Research	19
Nephrology Dialysis Transplantation	101
Protein Engineering Design and Selection	3
Physiological Genomics	9
Rheumatology	122
RNA	1
Toxicological Sciences	21

The MEDLINE files are in the protected area of the Web site as follows:

MEDLINE_ASCII.gz - 132 MB, expands to a single file of about 500 MB
MEDLINE_XML.gz - 160 MB, expands to a single file of about 1.5 GB
Error_pmids.txt - list of the erroneous 1,767 PMIDs
Pmids.txt - list of all 162K PMIDs

The same restrictions apply to the NLM data as to the Highwire data: If you have signed the TREC usage agreement, you are free to use the data in any way, but it must not be posted on any public Web site.

Another file in the protected area of the Web site is legalspans.txt. This file contains all "legal spans" for all documents in the collection. Legal spans are defined as any text >0 bytes in length between an HTML paragraph tag, which is any tag that starts with <p or </p. There are a total of 12,641,127 legal spans in the collection. We will use these spans to define allowed passages in the pooling and evaluation process, and you may find them useful in other ways. The file has the format:

pmid<tab>span-start<tab>span-length

Note that the first and last span (and some in between) may contain a great deal of text that would not be a true passage, e.g., HTML <head> and Javascript. Here is an example of the legalspans.txt file from document 15485830.html:

15485830	0	403
15485830	406	133
15485830	542	6
15485830	551	88
15485830	642	3660
15485830	4305	3828
15485830	8136	16
15485830	8155	2923
15485830	11081	1486
15485830	12570	2453
15485830	15026	1548
15485830	16577	3237
15485830	19817	814
15485830	20634	829
15485830	21466	701
15485830	22170	989
15485830	23162	1433
15485830	24598	840
15485830	25441	4191
15485830	29635	16
15485830	29654	3347
15485830	33004	81
15485830	33088	1961
15485830	35052	1949
15485830	37004	1745
15485830	38752	16
15485830	38771	1469
15485830	40243	2338
15485830	42584	16
15485830	42603	4528
15485830	47134	1191
15485830	48328	786
15485830	49117	2549
15485830	51669	1945
15485830	53617	16
15485830	53636	1607
15485830	55246	2798
15485830	58047	16
15485830	58066	2157
15485830	60226	3324
15485830	63553	3722
15485830	67278	3022
15485830	70303	2977
15485830	73283	1808
15485830	75094	734
15485830	75831	1036
15485830	76870	603
15485830	77476	831
15485830	78310	529
15485830	78842	20417
15485830	99262	16

Topics

There are 36 official topics, which are listed in a text file in the protected area of the Web site.

Relevance Judgments

The gold standard correct passages, along with topicid and aspect MeSH terms, are in a text file in the protected area of the Web site and are in the following format:

<topicid> <pubmedid> <offset> <length> <MeSH aspects, separated by '|' characters>

A text file of all judgments is available as well. Unlike the previous file, this file lists each passage that was judged and whether any part of it was relevant or not.

Evaluation Tools

The program to do the evaluation is a slightly updated version of the 2006 Python script.