The Python project 'DNNGP' can be used to implement genome-wide prediction (GP), which can predict the phenotypes of plants and animals based on multi-omics data. The code is written using Python 3.6 and TensorFlow 1.15.
- Version 1.0 -First version released on August, 20th, 2022
DNNGP requires Python 3.6. Follow the instructions at https://www.tensorflow.org/install/gpu to set up GPU support for faster model training. Once GPU is set up, install with conda by executing these instructions from the root of the checked-out repository:
conda create -n dnngp python=3.6.5
conda activate dnngp
pip install -r requirements.txt
Users can also use DNNGP on CPU, and the installation method is the same as above.
Download the release package and unzip to your working directory.
To run locally, there are two required input files. One file contains the phenotype of interest, the other file contains the SNP data, genomic expression data or other related omics data with digital coding.
An example command to train DNNGP to predict the phenotype pheno from the SNP data, genomic expression data or other related omics data with digital coding is the following:
python dnngp_runner.py \
--batch_size [num] \
--epoch [num] \
--lr [num] \
--patience [num] \
--dropout1 [num] \
--dropout2 [num] \
--SNP [your omics-data file] \
--pheno [your phenotype data file]
Of particular note is the run.py
. This script is used to get results in batches by adjusting different hyperparameters and inputs.
The example-data files are alread included in the release package,
You can also get the exmample data from source code.
Download example-data.tgz and extract data with tar zxvf example-data.tgz
.
Kelin Wang๏ผ[email protected]๏ผ
Huihui Li๏ผ[email protected]๏ผ
GPLv3 ยฉ Kelin Wang, Huihui Li