Abstract
Phylogenetic methods have shown great promise in understanding the development of broadly neutralizing antibody lineages (bNAbs). However, mutational process for generating these lineages - somatic hypermutation (SHM) - is biased by hotspot motifs, which violates important assumptions in most phylogenetic substitution models. Here, we develop a modified GY94-type substitution model which partially accounts for this context-dependency while preserving independence of sites in calculation. This model shows a substantially better fit to three well-characterized bNAb lineages than the standard GY94 model. We show through simulations that accounting for this can lead to reduced bias of other substitution parameters, and more accurate ancestral state reconstructions. We further explore other implications of this model; namely, that the number of hotspot motifs - and therefore likely the mutation rate in general - is expected to decay over time in individual bNAb lineages.
