| Abstract | Biological information connects evolutionary dynamics with the tools of information theory by focusing on the distribution of alleles at each locus. At a neutral locus all possible alleles are equally likely in the long-term, resulting in a maximum entropy (minimum information) distribution. For loci with fitness effects, a preference for particular alleles, created by mutation and natural selection, results in less entropy and more information. Previous research has primarily treated loci as non-interacting by making specific approximations in the calculation of biological information. In the present research, we refine previous approximations to include epistasis or interactions between loci. To understand the results of the new approximation, we consider evolutionary dynamics using the discrete time quasispecies equation. This model describes mutation and selection in an infinite population, removing the need for statistical inference in the current investigation. Application of the improved form for the biological information demonstrates that mutual information between loci is only present when considering evolution on epistatic fitness landscapes. Examples of two-locus, two-allele fitness landscapes with and without epistatic interactions will be presented. Finally, we consider modular four-locus, two-allele fitness landscapes. As modular interactions are inherently epistatic, we demonstrate that our refined approximation provides great insight into the underlying structure of the fitness landscape. |
|---|
