Finding General Patterns in Fitness Landscapes

Abstract

Biological phenomena can be examined at multiple levels of organization. For example, the role of individual amino acids in protein function representing a low level, the interactions between genes underlying complex traits as an intermediate level, and evolutionary processes across taxa at the highest level. Phenomena at each level must emerge from processes at lower levels. Thus, understanding this emergence is crucial for a full understanding of any biological phenomenon, including the genetic basis of disease and the course of evolution. My thesis focuses on how mutational effects propagate across these levels of organization.

My second chapter develops a null theory for how mutational effects scale to complex phenotypes, from which universal evolutionary patterns emerge. Although mutational effects have predictable, deterministic effects on lower level phenotypes, the complexity of interactions that determine phenotypes like fitness results in a seeming randomness of mutational effects, or idiosyncrasy, at this higher level. Universal evolutionary patterns in adaptation, mutation accumulation, and the effects of mutations across fitness levels then emerge as statistical laws from this randomness. This null theory of “idiosyncratic interactions” resolves a paradox in the common, but mistaken, interpretation of these universal evolutionary patterns. My third and fourth chapters capture a deviation from this null theory, showing that a feature at the lowest level, the radical- or conservative-ness of amino acid substitutions, leads to predictable fitness differences and may bias phylogenetic patterns. Across most phyla, there is a substitution bias towards transitions versus transversions; there has been a long-standing debate over whether this bias is due to a mutational bias or selection. In my third chapter I show that transversions are more detrimental to fitness in two RNA viruses, influenza virus and HIV. This can be partly explained by their greater likelihood to cause radical amino acid changes, which are more detrimental. Thus, selection is likely a major contributor to the transition-transversion substation bias. My fourth chapter examines further consequences of radical versus conservative amino acid changes. Uncovering the molecular constraints on proteins and the changes responsible for adaptations are major goals in molecular evolution. I find that while proteins are constrained in accepting radical amino acid changes in influenza virus, HIV, and additionally in Zika virus, such radical changes are also the most beneficial, conditional on being beneficial. Multi-nucleotide mutations, which are more likely to cause radical changes due to the genetic code, show the same pattern as radical changes. These findings have implications for viral evolution, phylogenetic inference of selection, and the evolution of the genetic code. Our null theory of idiosyncratic epistasis may help guide future work on how deviations from our predictions at the level of evolutionary patterns reflect the organization of interactions at the level of complex phenotypes and the biological properties of individual genes or mutations at the lowest level.