Figure 3 | Scientific Reports

Figure 3

From: Using human in vitro transcriptome analysis to build trustworthy machine learning models for prediction of animal drug toxicity

Figure 3

Investigation into feature importance for ML model. Showing the results from (a) the ExtraTreesRegressor approach, trained on L1000 gene expression data and chemical structure information. Showing feature importance as a bar chart for the 964 landmark genes and 166-bit MACCS chemical fingerprints, alongside a cumulative line plot of feature importance and (b) the highest feature contributions (>0.5) for the gene expression-based dimension with the shortest length scale in our best GP model.

Back to article page