TREC 2005 Genomics Track Data

This page lists the files that are in the distribution for the TREC Genomics Track. More detail about these files can be found in the 2005 track protocol. The data files themselves are located in the active user portion of the track Web site at http://ir.ohsu.edu/genomics/data/. This area is password protected, with the password only available to those who have completed data usage agreements and/or are registered for TREC. Please note that the data in the active user portion of the NIST TREC Web site are out of date and incomplete.

Ad Hoc Retrieval Task

Documents

MEDLINE subset - available from 2004 track data

Topics

Official topics

adhoc2005narrative.txt - Narrative version of official topics (100-149)
adhoc2005topics.doc - Tabular (Microsoft Word) version of official topics (100-149)
adhoc2005topics.pdf - Tabular (PDF) version of official topics (100-149)

Sample topics

90.txt - 99.txt - Sample searches and relevance judgments for sample topics (see caveats on protocol page)

Relevance judgments

Relevant documents for scoring runs using trec_eval

genomics.qrels.large.txt - file of all relevance judgments done, with "0" indicating not relevant, "1" possibly relevant, and "2" definitely relevant (NOTE: These are the official qrels for 2005 and must be used with trec_eval to insure that the bpref measure is calculated properly.)
genomics.qrels.txt - smaller file listing relevant documents only, with definitely and possibly relevants designated by "1"

Categorization Task

Documents

Highwire training and test documents - available from 2004 track data

PMID-full text crosswalk

For training and test documents (these are identical to the files packed in the gzip archive for the 2004 data but are made available here for convenience):

train.crosswalk.txt
test.crosswalk.txt

Sample, training, and test files

File contents	Sample output file	Training data file	Test data file
A (alelle)	sample.Atrain.txt (18 KB)	Atrain.txt (4 KB)	Atest.txt (4 KB)
E (expression)	sample.Etrain.txt (64 KB)	Etrain.txt (1 KB)	Etest.txt (2 KB)
G (GO annotation)	sample.Gtrain.txt (28 KB)	Gtrain.txt (5 KB)	Gtest.txt (6 KB)
T (tumor)	sample.Ttrain.txt (16 KB)	Ttrain.txt (1 KB)	Ttest.txt (1 KB)

Gene tagging of MEDLINE corpus

2004_TREC_MEDLINE_1_MGI_FOR_PMID.tsv.txt - automated mouse gene tagging of first half of ad hoc MEDLINE corpus
2004_TREC_MEDLINE_2_MGI_FOR_PMID.tsv.txt - automated mouse gene tagging of second half of ad hoc MEDLINE corpus

Evaluation program cat_eval

The required version of the program is 2.0, updated on Sept. 9, 2005. Sample data to test the program are provided as described in the above table. The program is provided in source code and as a Windows executable (see protocol page for documentation):

cat_eval2.py - For this version, you need Python 2.3 or above installed. Install as you would any script
cat_eval2.zip - For this version, unzip to a directory on a Windows machine and execute cat_eval2.exe; no Python installation is necessary

Last update - September 30, 2005