Fig. 2

The sequence identity between repeated units can bias the inference of evolutionary couplings. Repeat sequences of the ANK familywere concatenated in a MSA of size 2L ₀=66 positions and ≈73000 sequences and co-variations were measured with direct information metric.a Sequence identity distributions between consecutive ANK repeats found in (x) natural proteins and (o) randomized pairs of repeats. b Direct information matrices between positions obtained without correcting (DI, upper half) or with proper equalization for repeat identity (DI _id , lower half)