Table 1:

Investigation of the expected antimicrobial properties of samples generated by AMPGAN v1 and v2 using the machine learning models developed by Waghu et al.. 5000 AMP candidates were drawn from each generative model and each candidate was evaluated by four predictive models: a support vector machine, a random forest, an artificial neural network, and discriminant analysis. The percentage of generated samples that were predicted to have antimicrobial activity is presented, along with a bootstrapped 95% confidence interval in parenthesis.

	AMPGAN v1	AMPGAN v2
Support Vector Machine	5.24% (4.44%, 6.08%)	79.85% (78.27%, 81.39%)
Random Forest	7.66% (6.68%, 8.72%)	88.36% (87.06%, 89.62%)
Artificial Neural Network	4.22% (3.52%, 5.00%)	88.24% (86.94%, 89.46%)
Discriminant Analysis	7.76% (6.76%, 8.72%)	83.71% (82.23%, 85.18%)