This Python package provides a robust and efficient solution to predict an individual's genetic ancestry using genomic data. The tool leverages federated learning to train a single model on separate datasets from different nodes.
- Federated Prediction of Ancestry: It can train a federated model using separated datasets on the different nodes. Individual genotypes are not shared between nodes.
- Global Ancestry Prediction: It uses global ancestry data to provide a broad overview of an individual's ancestral roots.
- High Performance: Designed to handle large genomic datasets effectively.
- User-Friendly: It provides a straightforward command-line interface, making it easy to use for both bioinformaticians and geneticists.
You can install the Ancestry Prediction tool using poetry. Make sure that you have Python 3.9 or later and poetry installed on your system.
poetry add https://github.com/genxnetwork/flan/
To fit the model globally, use the following command:
flan global fit
After fitting the model, you can predict the global ancestry using the following command. Replace file.vcf.gz with the path to your own vcf file.
flan global predict --file=file.vcf.gz
To fit the model using the client-server federated architecture, you need to launch server which will perform the aggregation. Server should be accessible from all client nodes.
Server side:
flan server fit --server.host=127.0.0.1 --server.port=57632
Then, you should launch clients on each node.
Client side:
flan client fit --server.host=127.0.0.1 --server.port=57632
You can then predict the ancestry on the client side using the federated model and the following command:
flan client predict --file=file.vcf.gz
If you use flan in your research, please cite it as follows:
TO BE ADDED BY THE PACKAGE AUTHOR
Contributions to improve this tool are welcome! You can contribute in several ways:
- Report bugs or suggest new features by opening an issue on our GitHub page.
- Improve the code: If you have a bug fix or a new feature you'd like to add, please open a pull request.
- Improve the documentation: If you find any mistake or ambiguity in the documentation, please make a pull request to help us improve it.