The FiLter for Understanding True Events (FLUTE) tool seeks to identify high-confidence signaling events from biomedical NLP output by comparing with existing biological databases. As such, FLUTE can reliably determine the confidence in the biomolecular events extracted by NLP methods and at the same time provide a speedup in event filtering by three orders of magnitude.
Understanding disease at the cellular level requires detailed knowledge of signaling networks. To aid in this task, many advances have been made in the field of natural language processing (NLP) to extract signaling events from biomedical literature. However, even state-of-the-art NLP methods incorrectly interpret some signaling events described in the literature.
- Unzip the
flute.sql.zip
file. - Log in to the MySQL environment using your username and password.
- From there, create an empty database named
YOUR_DATABASE_NAME
(any name is fine, I recommend the classicflute
). - Log back out, and again from the command line
mysql -u YOUR_USER_NAME -p YOUR_DATABASE_NAME < flute.sql
- If you created a username and password, this will be your username in the above command, but do not enter your password above! Once you hit enter, it will prompt you for the password.
- You can now run the
run_FLUTE.py
script, you will need to enter the database, host, username, etc. as an argument from the command line.
To filter interactions, run run_FLUTE.py
. You must have Python3 installed. The script takes several parameters:
- MySQL username
- MySQL password
- Host name -
localhost
for macOS/Linux, desktop name for Windows - Database name (
YOUR_DATABASE_NAME
) - Input filename
- Output filename for interactions
- Output filename for scores
Input files must have the following headers: RegulatedName, RegulatedID, RegulatedType, RegulatorName, RegulatorID, RegulatorType, PaperID
. Output files include list of reading interactions that pass filtration, and the filtration scores for those filtered interactions.
For additional FLUTE functionality beyond interaction filtering (get related papers, related interactions in databases, etc.), check python -h
- Download the appropriate distribution for your OS from here
- Restart your computer, and add to path if necessary.
- From the command line, access the MySQL environment by typing:
If the first prompt fails, you may need to enter the password associated with your computer user account:
mysql -u root
mysql -u root -p
- You may choose to create a local username and password to keep your database private. However, this is published information, so it is not a security risk if you do not have a password.
- Install MySQL Python connector
Emilee Holtzapple, Cheryl A Telmer, Natasa Miskov-Zivanov, FLUTE: Fast and reliable knowledge retrieval from biomedical literature, Database, Volume 2020, 2020, https://doi.org/10.1093/database/baaa056
This work was funded in part by DARPA Big Mechanism award, AIMCancer (W911NF-17-1-0135); and in part by the University of Pittsburgh, Swanson School of Engineering.
To be updated