Outputs

1. Visualize AUC results for the cumulative distribution curves in MLflow:

Connect to MLflow:
mlflow ui -p 5000
Open internet browser on the following address:http://localhost:5000/

First MLflow window opening shows the models run:

To visualize the cumulative distribution curve, select the relevant models and click on compare:

On the bottom left of the opening page, click on cum_dist:

Click on step from the choices on the left for a visualization of all protein coding genes on the x axis:

Plots can be downloaded, and more information is revealed by placing the pointer on the curves, such as the percentage of known disease genes found in the first 16 predictions:

All results can be found in the /reports folder:

2. Model performances and parameters:

Model performances based on the AUC of the cumulative curve, with all the run parameters per model can be found in files “loocv_best-runs-by-model” for LOOCV results and in files “test_best-runs-by-model” for the test run results.

3. Gene rankings:

Ranking results of the LOOCV of the unlabeled genes can be found in the file “predictions”.

The predictions of each model on each node is presented in a heatmap:

4. Statistical comparisons:

Results for the statistical pairwise metrics can be found in figure format in the files “loocv_best-runs-by-model_pairplot.png” or “test_best-runs-by-model_pairplot.png”.