Table 1 Micro- and macro-averaged F1 scores for all models across EC levels 1-4 on the test set

From: Interpretable Kolmogorov-Arnold networks for enzyme commission number prediction

Metric

Level

CLEAN

DeepECtransformer

DeepEC

  

MLP

KAN

MLP

KAN

MLP

KAN

Micro F1-scores

1

0.968 ± 0.007

0.971 ± 0.008*

0.784 ± 0.059

0.905 ± 0.052*

0.895 ± 0.019

0.915 ± 0.014*

2

0.938 ± 0.012

0.948 ± 0.011*

0.773 ± 0.063

0.894 ± 0.059*

0.887 ± 0.022

0.914 ± 0.019*

3

0.882 ± 0.014

0.891 ± 0.013*

0.768 ± 0.064

0.880 ± 0.060*

0.880 ± 0.019

0.890 ± 0.018*

4

0.875 ± 0.010

0.880 ± 0.010*

0.787 ± 0.015

0.867 ± 0.046*

0.859 ± 0.014

0.874 ± 0.013*

Macro F1-scores

1

0.962 ± 0.021

0.965 ± 0.019*

0.797 ± 0.057

0.903 ± 0.051*

0.906 ± 0.028

0.923 ± 0.020*

2

0.910 ± 0.013

0.915 ± 0.010*

0.687 ± 0.086

0.819 ± 0.071*

0.808 ± 0.021

0.899 ± 0.019*

3

0.809 ± 0.018

0.812 ± 0.012*

0.582 ± 0.095

0.781 ± 0.080*

0.720 ± 0.022

0.791 ± 0.014*

4

0.802 ± 0.024

0.811 ± 0.021*

0.585 ± 0.068

0.727 ± 0.059*

0.589 ± 0.019

0.737 ± 0.013*

  1. Asterisks (*) denote statistical significance by the Wilcoxon signed-rank test; boldface indicates the best performing variant (MLP or KAN) within each model.
Back to article page