DeepGS is a deep learning framework that aims to predict phenotypes from genoypes.
The main features:
DeepGS is an R package covering GS cross-validation experiment design, evaluation and combination of different GS predictions. It generates comprehensive results for candidate population phenotypic values, GS prediction performance evaluation, and ensemble predicted phenotypic values among different GS models.
DeepGS also takes advantage of machine learning technologies for GS prediction performance, using deep convolutional neural network, and ensemble of algorithm for integrating different GS models predictions, using particle swarm optimization.
DeepGS has been successfully applied to predict eight phenotypic traits on a population of 2,000 Iranian bread wheat (Triticum aestivum) lines from the wheat gene bank of the International Maize and Wheat Improvement Center (CIMMYT). Experimental results demonstrated that the DeepGS and ensemble models achieved competitive prediction performance for high phenotypic individuals based on mean normalized discounted cumulative gain value (MNV).
DeepGS is developed and maintained by the lab of Prof. Chuang Ma at Northwest A&F University. For comments/suggestions/error reports, please contact Wenlong Ma ([email protected]) or Chuang Ma ([email protected])