Winners Announced for NLP Challenge to Harness the Power of Biomedical Data

LITCOIN Challenge Graphic

Image Credit: National Center for Advancing Translational Sciences

The number of scientific studies is constantly increasing, but a large amount of biomedical information from these studies is not easily accessible to the researchers and data scientists who can use it to improve health. Simply placing biomedical data in open data repositories is not enough to deliver public health solutions. Many of these repositories are either incomplete or too large to be fully explored.

The LitCoin Natural Language Processing (NLP) Challenge aimed to address this accessibility issue and increase the power and value of data to identify new treatments. Through the challenge, NCATS rewarded the most creative and effective uses of free text from biomedical publications to create knowledge graphs that relate concepts within existing research. By engaging technologists, the scientific and medical community and the public, we can help researchers find connections that otherwise may have been difficult to discover, drastically increasing the data’s value in solving public health issues. The LitCoin challenge is part of a broader initiative at NCATS to change the “currency” of biomedical research.

NCATS partnered with the NASA Tournament Lab to find the ideal challenge platform, selecting CrowdPlat and the artificial intelligence competition platform bitgrit. Participants used information from published abstracts to create NLP systems that could recognize biomedical concepts and relationships between them. Using a custom evaluator program, the output of each NLP system was compared to a set of hand-annotated assertions created by researchers at the National Center for Biotechnology Information. Judges used the scores from the evaluator program to determine the winning submissions, subject to the final decision by the Award Approving Official.

The software packages built for this challenge will be used to further NCATS’ goal of creating a new type of biomedical publication. This will encourage the sharing of computationally accessible data from the time of publication with the goal of changing the way research results are compiled for use when developing public health policies and new treatments and cures. The results of the challenge showcase what is possible when scientists and computational researchers come together to change the research landscape.


1st place: JZhangLab@FSU (Florida State University, Tallahassee, Florida)

2nd place: UTHealth SBMI (University of Texas Health Science Center, Houston, Texas)

3rd place: UIUCBioNLP (University of Illinois at Urbana-Champaign, Champaign, Illinois)

Runners Up:

  • OsborneLabUAB (University of Alabama at Birmingham, Birmingham, Alabama)
  • RMIT-READBio (RMIT University, Melbourne, Australia)
  • UMaas (Maastricht University, Maastricht, Netherlands)
  • LasigeUnicage (University of Lisbon, Lisbon, Portugal)

Honorable Mention: Hao Liu (Columbia University Irving Medical Center, New York, New York)