Indeed, one of the paleo-Americans, whom Chatters seemed not to want to discuss, was tucked away all by himself in a far corner of the graph, like a wallflower.
In ROC space, a classifier has strong predictive value if its points lie in the upper, left-hand corner of the graph, representing an FPR close to 0 and a TPR close to 1.
We can see that mFPT has its ROC curve climbing rapidly towards upper left hand corner of the graph, revealing a good prediction performance.
In contrast, the 1918 pandemic was positioned at the upper right corner of the graph, indicating a very transmissible and clinically severe pandemic with extensive effects in the population.
FPR is a measure of how many of the known non-interactions are incorrectly predicted as interactions and is defined as F P F P + T N. In PR space, a classifier has strong predictive value if its points lie in the upper, right-hand corner of the graph, representing a precision and recall close to 1.
In order to maximize the TP rate while simultaneously minimizing the FP rate, one should select the curve that is the closest to the upper-left corner of the graph (indicated by a circle in Figure 3).
The performance of an algorithm increases when its associated trade-off points are closer to the left top corner of each graph (higher DP with a lower FAR).
In Figure 2a, there are several values deviating from a straight line in the lower left corner of the graph, and the line is bent slightly.
The similar, small secretome-to-proteome size ratios of the Saccharomycotina are clearly seen in the form of a "dip" of the data points in the lower left corner of the graph, while for the other species, these show more variation.
For 8 among 22 tRNAs (36%), those in the lower left corner of the graph, it does not matter much whether one uses a row- or column-based approach.
