Fig. 5.

The difference between two gain functions G^(2,1)(θ, y) and G^(2,2)(θ, y). Each row indicates a configuration of θ on the target sequence x and each homologous sequence x′, and its positive contributions to the gain functions for y_ij = 1. For example, the middle row means that the configuration of θ_ij^x = 0 and θ_ik^xx′ = θ_kl^x′ = θ_jl^xx′ = 1 have a positive (1 − α) contribution for y_ij = 1 in the gain function G^(2,1), while have no contribution in the gain function G^(2,2).