Table 3 Hotspot model selection
Model nameConstraint/optimization of each haP-values from LR tests
hWRChGYWhWAhTWhSYChGRSCH103VRC26VRC01
Symmetric WRC/GYWaMLhWRC00004.6 × 10−156.3 × 10−055.2 × 10−03
Asymmetric WRC/GYWMLML0000
Symmetric WA/TWa00MLhWA00<1 × 10−15<1 × 10−15<1 × 10−15
Asymmetric WA/TW00MLML00
Symmetric SYC/GRSa0000MLhSYC0.532.0 × 10−134.2 × 10−03
Asymmetric SYC/GRS0000MLML
Uniform hotspotsaMLhWRChWRChWRC004.2 × 10−15<1 × 10−15<1 × 10−15
Hierarchical hotspotsMLhWRCMLhWA00
SCAHaMLMLMLMLMLhSYC0.651.2 × 10−069.1 × 10−04
FCHMLMLMLMLMLML
  • Models of hotspot hierarchy (degree of mutability) and symmetry, specified by placing constraints on how different values of h are optimized. Columns 2–7 show how the parameter ha is obtained for a particular model. A value of 0 indicates that h is fixed at zero, ML indicates that a parameter is optimized by ML, and ha indicates that h parameter is equal to another value of h. For instance, in Symmetric WRC/GYW, hGYW is equal to its reverse complement hWRC, which is ML optimized. However, in Asymmetric WRC/GYW, both are ML optimized. Rows 8–10 show P-values obtained from likelihood-ratio tests of each of these nested hotspot models for the bNAb lineage specified in each column. Parameters, log likelihood, and AIC of each fit are shown in Figure S3 in File S1. LR, likelihood ratio; SCAH, symmetric coldspots, asymmetric hotspots; FCH, free coldspots and hotspots.

  • a Each of these models is nested with the model immediately below it by one free parameter, allowing hypothesis testing using a likelihood-ratio test.