Skip to main content

Table 3 Comparison of our machine learning method and Flux Balance Analyses on glucose minimal media condition

From: Machine learning based analyses on metabolic networks supports high-throughput knockout screens

Performance ML\BFV1 ML2 FBA3
true positives 192 266 174
true negatives 932 968 971
false positives 64 28 25
false negatives 146 72 164
sensitivity 56.80% 78.70% 51.48%
specificity 93.57% 97.19% 97.49%
positive predictive values 75.00% 90.48% 87.44%
negative predictive values 86.46% 93.08% 85.55%
overall accuracy 84.26% 92.50% 85.83%
  1. 1machine learning without the feature BFV (biomass flux value from the FBA).
  2. 2machine learning including the feature BFV
  3. 3Flux Balance Analysis