/DL-Biobank

Can deep learning improve genomic prediction of complex human traits? by Pau Bellot, Gustavo de los Campos, Miguel Pérez-Enciso, Genetics

Primary LanguagePythonMIT LicenseMIT

Can deep learning improve genomic prediction of complex human traits? by Pau Bellot, Gustavo de los Campos, Miguel Pérez-Enciso, Genetics.

This repo contains the code to reproduce the experiments of the paper Can deep learning improve genomic prediction of complex human traits?.

Before you start

  • Download and install Tensorflow and Keras.
    • The file requirements.txt contains a simple list of all the packages in the current environment, and their respective versions.
  • Download the biobank dataset and traits.
    • Save genotypes
    • Save phenotypes in a csv, each trait should be in a column
  • Run a GWAS for all traits and save the results in a csv file
    • This csv should have the p-values of each SNP for each trait (in columns)

data

  • data/ folder contains the GWAS for all the traits used in the paper

To run

Train MLPs models for height with 10k Best SNPS

python main.py --mlp --trait height

Train CNNs models for height with 10k Unif SNPS

python main.py --cnn --trait height --unif

Train MLPs models for BHMD with 50k Best SNPS

python main.py --mlp --trait BHMD --unif -k 50000

Deep learning hyperparamer tuning

  • see GA directory

To generate the figures eneration

  • see notebook and R code

License

MIT