An experimental program created at the 9th Microbes & Food Safety Bioinformatics Hackathon to summarise the literature for a particular gene using Chat-GPT
The algorithm searches PubMed for literature on a particular gene within a species. It then uses Chat-GPT to generate summaries in a hierarchical manner first for each paragraph, then each paper and finally for the set of papers.
The program requires OpenAI and Google search API's that must be supplied by the user.
The easiest way to install the program is to clone the GitHub repository and install using pip. To include the API's create a file called api_keys.yaml and place it in the folder genesearch/genesearch/. The file should be formatted as
google_api_key: "your google API key"
google_engine_id: "your google search engine ID"
openai_api_key: "your openAI API key"
Alternatively you can set environment variables by running
export GOOGLE_API_KEY="your google API key"
export GOOGLE_ENGINE_ID="your google search engine ID"
export OPENAI_API_KEY="your openAI API key"
The program can then be installed by running
cd genesearch
pip install .
The OpenAI API key can be obtained by signing up for an account here.
A custom Google search API can be set up by following the instructions here and setting up a custom search engine using the control panel here. At the moment we have tested the program by specifying ncbi.nlm.nih.gov/* in the "sites to search" section.
After installation the program can be run by specifying the gene and species as
genesearch -g "gene name" -s "species name"
usage: genesearch [-h] -g GENE -s SPECIES [-n NUMBER_PAPERS]
[--max-paragraphs MAX_PARA]
[-l {DEBUG,INFO,WARNING,ERROR,CRITICAL}] [--version]
genesearch: A tool to summarise the literature for a particluar gene in a
species.
optional arguments:
-h, --help show this help message and exit
-l {DEBUG,INFO,WARNING,ERROR,CRITICAL}, --loglevel {DEBUG,INFO,WARNING,ERROR,CRITICAL}
Set the logging threshold.
--version show program's version number and exit
Input/output:
-g GENE, --gene GENE The gene name.
-s SPECIES, --species SPECIES
The species name.
-n NUMBER_PAPERS, --num-papers NUMBER_PAPERS
Number of papers to summarise (default=3)
--max-paragraphs MAX_PARA
Maximum number of paragraphs to include per paper
(default=10)