AIBreeding/XAI

Dataset

• For the present study, we used data from the GxE competition advocated by the G2F project in 2022 (https://www.maizegxeprediction2022.org/), including genetic markers (G2F-G) for maize inbred lines, phenotypic measurements (G2F-P) collected throughout each growing season, metadata (G2F-M) for each field trial, environmental covariate (EC) data, and environmental (G2F-E) data. G2F-E data were mainly climatic and soil variables captured during crop development in each experimental trial.
  

• In order to explore the influence of environmental factors on yield prediction results, we designed two sets of prediction scenarios: 1) yield prediction based on the whole genome, and 2) yield prediction integrating genome, weather and soil factors. Different data sets are generated for different prediction scenarios.
  

• For a detailed description of this dataset, please refer to the methods section of the paper.

Dataset file structure directory

├─test_set
│      New_test_values.csv
│      test_G.csv
│      test_GE.csv
│
└─train_set
       G.csv
       GE.csv
       New_Yield_values.csv
       train_Yield_folds.csv

Description

train_set

• G.csv
  Genome-wide principal component data used to train the G2P model.
• GE.csv
  The data was integrated from genome-wide principal component data, weather data and soil data.
• train_Yield_folds.csv
  The dataset is a ten-fold cross-validation dataset generated by the kfolds.py script for model training and testing.
  
• New_Yield_values.csv This dataset is assembled from the base model predictions and is primarily used to train the second layer of models in the stacking framework.
  

test_set

• test_G.csv This dataset is a predicted population of target genotypes from an untested environment and is used to validate the predictive performance of the model when environmental effects are ignored.
  
• test_GE.csv This dataset was integrated from genotype and environment to validate the predictive performance of the model across environments under environmental stress.
  
• New_test_values.csv This dataset is composed of the predicted values of the base model in the new environment and is used as a prediction set for the second layer of the model in the stacking framework.