Skip to main content
. 2019 May 2;15(5):e8747. doi: 10.15252/msb.20188747

Figure 2. Protection from challenge can be robustly modeled with antibody profiles.

Figure 2

  1. The predicted survival probabilities in the final model closely match observed KM curves for the IL‐12 adjuvanted group (red) and for the others (blue). Log‐rank tests indicate insignificant difference between predicted (solid) and observed (dashed) curves. n = 8 for IL‐12 and n = 24 for Others.
  2. The predicted relative risk of infection for each animal in the representative eightfold cross‐validation run (relative to mean at 0, horizontal dashed line) closely matches observed challenge data (concordance (C)‐index). The colors represent the adjuvant groups (groupID) as shown in the central legend box to the right side of panel (D).
  3. Animals in the IL‐12 adjuvanted group (red) have significantly lower predicted risk of infection than those in the combined other groups (blues) in the representative eightfold cross‐validation run (Wilcoxon–Mann–Whitney). n = 8 for IL‐12 and n = 24 for Others. The colors represent the adjuvant groups (groupID) as shown in the central legend box to the right side of panel (D). The horizontal line in the box represents the median, the upper and lower limits represent the 3rd and 1st quartiles respectively, and the whiskers extends from the upper/lower limit to the highest/lowest value that is within 1.5 * (interquartile range) of the limit.
  4. The approach is robust, with models trained on actual data consistently obtaining high C‐indices (100 repetitions of eightfold cross‐validation yield mean 0.73 ± 0.02) and significantly (tail probability) and substantially (Cliff's Δ) outperforming those trained on permuted data (0.61 ± 0.08) and baseline C‐index for random prediction (0.5).
  5. A small set of features (columns) contribute to the final model (coefficients in bars; top panel), with one predictive of risk and three of protection (Cox PH P‐values: **P < 0.01; *P < 0.05; ‐, not significant). The individual animals (rows) are colored by adjuvant group and ordered in ascending order of time‐to‐infection.
  6. Substitution analysis reveals co‐correlates of the features from the final model (E), dominated by Gag specificities as well as ability to bind C1q.