FIGURE 1.

Illustration of the evaluation of clustering algorithms. Accuracy and coverage metrics were used to analyze clustering algorithms. Individually, these metrics do not necessarily indicate a good clustering algorithm, instead a trade-off between accuracy and coverage should be monitored. (A) High accuracy is achieved by making small clusters. These are likely to be epitope-specific; however, most antibodies are not contained in a cluster. (B) High coverage is achieved by the formation of large clusters. These contain most of the antibodies in the dataset but do not tend to be epitope-specific.