Skip to main content
[Preprint]. 2023 Sep 19:2023.07.10.548341. Originally published 2023 Jul 11. [Version 2] doi: 10.1101/2023.07.10.548341

Figure 3. A. The CASP15 assembly groups,

Figure 3.

ordered from left to right according to their relative performance. For the final ranking, we utilized the average scores obtained from the jackknifing protocol (Methods). The bar height represents the cumulative Z-score (Eq. 7), while the error whiskers mark the standard deviation of the jackknifed data set. Human groups are shown in blue, servers in green, and the baseline NBIS-AF2-Multimer in orange. B. Model ranking efficiency of the top performing groups. The ICS of the first submitted model is plotted against the best ICS model. Yellow boxes highlight off-diagonal targets (i.e., those, for which predictors wrongly identified their best model).