Fitness Decline Over Long-Term Evolution of a Small Population of Asexual Computational Organisms

Abstract

The vital information carried in the DNA of every organism must be protected from the mutagenic processes that tend to degrade it. Molecular systems that detect and correct chemical alterations and base-pair mismatches form the first line of defense in all kingdoms of life. Natural selection provides a second line of defense. Specifically, purifying selection (or negative selection) is the natural tendency in wild populations for the genetic lines of individuals that suffer impairment from a new mutation to terminate within a few generations of the mutation event. Purifying selection is known to be much more efficient than positive selection, though it becomes less efficient in very small populations. Here, we describe a long-term evolution experiment that tracks fitness in a population of 1,000 computational organisms. We use the previously described Stylus artificial-world model, in which genes encode drawings that have scorable functionality based on the degree to which they resemble one of the Chinese written characters. In our weakest-link minimal-genome model, the fitness of the organism is proportional to the lowest score of its 223 essential genes. Following a population of 1,000 model organisms through 2,000,000 fixation events, we find that fitness declines approximately as a two-phase exponential decay. The long-term result is substantial loss of function for all 223 genes, seen both in a collapse of numerical scores and in loss of legibility. The cause of the collapse is an imbalance in the initial genome: the number of ways for mutations to produce a new worst gene is so much higher than the number of ways to improve the current worst that selection is unable to prevent decline. This result raises the question of genome decay in real organisms. It seems likely that the same imbalance exists there, though life may have ways of averting the decline we describe here.