Jianzhi (George) Zhang is a Professor of Ecology and Evolutionary Biology interested in the relative roles of chance and necessity in evolution. He got his B. S. from Fudan University in Shanghai, China, and his Ph. D. in Genetics from Pennsylvania State University. He was a Fogarty postdoctoral fellow at the National Institute of Allergy and Infectious Diseases before moving to the University of Michigan.

Abstract

Adaptive evolution proceeds through beneficial nucleotide substitutions in the genome, yet the number of such substitutions (N) required to reach a local or global fitness peak where no single mutation further increases fitness remains unknown. Here we estimate N by simulating adaptive walks on two large, complete, multi-environment adaptive landscapes inferred from massive empirical data using machine learning, with validation from smaller experimentally mapped landscapes. We find that N rises linearly with the number of variable sites (L) in the landscape, regardless of prior adaptation in another environment. Extrapolation suggests a minimal N of 10^5 for typical prokaryotes and 10^7 for mammals. By contrast, in highly rugged shuffled landscapes, N is markedly reduced, while fitness gains also shrink. The relatively smooth empirical landscapes therefore enable greater fitness gains while lengthening adaptive walks, rendering even local fitness peaks effectively unreachable before environments change. These results explain the persistence of fitness gains in long-term evolution experiments and the widespread occurrence of beneficial mutations across species. They suggest that populations continuously adapt while remaining far from fitness optima, challenging the long-held view that adaptation culminates at fitness peaks.