Assessment of accuracy

Two methods were applied to assess the accuracy of the predictions. Average Q₃[19], and Segment Overlap[64].

Q₃ is a measure of the overall percentage of predicted residues, to observed:

$$Q_{3} = \sum_{(i=H,E,C)}\frac{predicted_{i}}{observed_{i}} \times 100$$ (1)

Segment overlap calculation[64] was performed for each data set. Segment overlap values attempt to capture segment prediction, and vary from an ignorance level of 37% (random protein pairs) to an average 90% level for homologous protein pairs. Segment overlap is calculated by:

$$Sov = \frac{1}{N} \sum_{s}\frac{minov (s_{obs};s_{pred})+\delta}{maxov (s_{obs};s_{pred})} \times len(s_{1})$$ (2)

Where N is the total number of residues, minov is the actual overlap, with maxov is the extent of the segment. $\delta$ is the accepted variation which assures a ratio of 1.0 where there are only minor deviations at the ends of segments[64].