TREC-PM (Precision Medicine)

NOTE: Maintenance of this project has moved to the JULIE Lab.

A repository containing support code and resources initially developed at the Institute for Medical Informatics, Statistics and Documentation at the Medical University of Graz (Austria) for participation at the 2017 TREC Precision Medicine Track. For further information on this track and the final results please check the official TREC-PM 2017 overview paper. Team name: imi_mug.

It was then further improved for participation at the 2018 TREC Precision Medicine Track. Improvements include: support for subtemplates and the possibility to use disjunctive queries (dis_max) allowing e.g. synonyms and hypernyms to have different weights. Team name: hpi-dhc.

Citing

If you use imi_mug's original data or code in your work, please cite their TREC 2017 proceedings paper:

TREC 2017 Precision Medicine - Medical University of Graz. Pablo López-García, Michel Oleynik, Zdenko Kasáč and Stefan Schulz. Text REtrieval Conference, Gaithersburg, MD. 2017. Available at https://trec.nist.gov/pubs/trec26/papers/imi_mug-PM.pdf.

If you use any of the improvements mentioned above, please also cite our TREC 2018 proceedings paper:

HPI-DHC at TREC 2018 Precision Medicine Track. Michel Oleynik, Erik Faessler, Ariane Morassi Sasso, et. al. Text REtrieval Conference, Gaithersburg, MD. 2018. Available at https://trec.nist.gov/pubs/trec27/papers/hpi-dhc-PM.pdf.

Other resources

2017

• imi_mug TREC 2017 presentation slides
• imi_mug TREC 2017 Poster
• TREC 2017 proceedings.

2018

• hpi_dhc TREC 2018 presentation slides
• hpi_dhc TREC 2018 Poster
• hpi_dhc TREC 2018 Data Artifacts
• TREC 2018 proceedings.

Code Dependencies

• JDK 8+
• maven
• make (for trec_eval tool)
• gcc (for trec_eval tool)
• perl (for sample_eval tool)
• Elasticsearch 5.4.0+

Some Examples on How to Run Experiments

# All executions should be run where the pom file is, usually the root of the project

# How to run the pubmed experimenter
# Necessary to define the year and type of gold-standard (for evaluation)

mvn clean install
mvn exec:java -Dexec.mainClass="at.medunigraz.imi.bst.trec.PubmedExperimenter"

# How to run the clinical trials experimenter
# Necessary to define the year and type of gold-standard (for evaluation)

mvn clean install
mvn exec:java -Dexec.mainClass="at.medunigraz.imi.bst.trec.ClinicalTrialsExperimenter"

# How to run the KeywordExperimenter
# Necessary to define the year and type of gold-standard (for evaluation)
# For positive booster, in the keyword template leave boost = 1
# For negative booster, in the keyword template leave boost = -1
# Also, in the KeywordExperimenter the keywordsSource needs to be specified

mvn clean install
mvn exec:java -Dexec.mainClass="at.medunigraz.imi.bst.trec.KeywordExperimenter" > out.txt &
cat out.txt | grep -e "\(^[0-9\.]*\)\(\;.*\)\(with.*\)\(\\[.*\\]\)\(.*\)" | sed -r "s/"\(^[0-9\.]*\)\(\;.*\)\(with.*\)\(\\[.*\\]\)\(.*\)"/\1 \2 \4/" > results.txt