Figure 2From: Machine learning based analyses on metabolic networks supports high-throughput knockout screensROC-curve showing our prediction results with different weight factors for positive instances. Each (blue) diamond shows the result for a different weight. From left to right, the weight was increased from 0.1 to 5.0, by a step size of 0.1. When the weight factors were higher than 1.0, the sensitivity remained constant. The dotted line was manually fitted.Back to article page