Selection can erase variation or protect it
The simplest way selection changes variation is by removing it. In Norway spruce, researchers describe both purifying selection, which eliminates harmful mutations, and positive selection, which favors beneficial ones, as forces that can reduce genetic diversity across the genome [4]. The same logic appears in humans: at the ACE2 gene, which the SARS-CoV-2 virus uses to enter cells, most of the 41 protein-altering variants found were rare in most populations, a pattern consistent with selection keeping potentially disruptive changes uncommon [1].
Yet selection does not always drain diversity. In fruit flies, sleep patterns differ systematically with latitude, and the gene-expression differences behind those patterns suggest natural selection is actively maintaining variation in circadian behavior rather than eliminating it [5]. The takeaway is that the direction depends on the kind of selection and the environment it acts in, not on selection being uniformly destructive.
When local conditions differ, selection can favor different variants in different places
A variant that is disadvantageous in one population may be useful in another, and that can keep multiple versions of a gene alive across a species' range. In the Drosophila sleep study, the latitudinal gradient in sleep behavior implies that different environments favor different genetic setups, which is one route by which selection preserves rather than erases variation [5].
The human ACE2 data show a similar geographic texture. Three protein-altering variants were common specifically among central African hunter-gatherers from Cameroon, at frequencies of roughly 8 to 16 percent, and sit on haplotypes showing signatures of positive selection [1]. Separately, selection signatures were detected in regulatory regions that influence how much ACE2 is produced, across multiple African populations [1]. The same gene carries very different variation patterns depending on where you look.
Hosts and pathogens chase each other, and that keeps variation in play
Some of the clearest cases involve infectious disease, where selection pressures are constantly renewed. In HPV16, a virus that causes more cancer than any other, researchers identified 56 viral codons putatively under positive selection for amino-acid change, and those codons disproportionately overlap the immune epitopes that the host's cytotoxic T cells recognize [2]. The interpretation offered is that a small subset of viral variants may improve persistence by escaping immune recognition, which would mean the host immune system is continuously selecting on viral variation [2].
The human side shows the same back-and-forth. The ACE2 variants under selection in African populations are associated with clinical phenotypes common in COVID-19 patients, and the authors suggest prior viral infections may have shaped this variation [1]. This is presented as a possibility, not a demonstrated cause, but it fits a pattern in which selection on host genes is driven by pathogens and leaves a detectable footprint in genetic variation.
The footprint selection leaves depends on the trait and the timescale
Selection's signature is not one thing. At TMPRSS2, another SARS-CoV-2-related host gene, researchers found 13 amino-acid changes that appear adaptive and specific to the human lineage compared with chimpanzees, a pattern of change fixed over long evolutionary time [1]. That is a different kind of evidence from a variant that is common in one population today, like the ACE2 variants in Cameroon [1].
The HPV16 work adds a third kind: codons under positive selection also overlap sites where viral subtypes have diverged, and show more rare variation in cervical precancers and cancers than in benign or cleared infections in the E1 protein [2]. Together these cases suggest selection's effect on variation is legible in several registers at once, from ancient fixed changes to ongoing population-level differences to variant frequencies that track disease outcome. What unites them is that selection is not a single force with a single result; it is a filter whose settings depend on the organism, the environment, and the pressure being applied.
Sources used in this answer
Impact of natural selection on global patterns of genetic variation and association with clinical phenotypes at genes involved in SARS-CoV-2 infection
At SARS-CoV-2 host genes, some ACE2 variants are common and positively selected in central African hunter-gatherers, selection signatures appear in ACE2 regulatory regions across multiple African populations, and 13 TMPRSS2 amino-acid changes appear adaptive and human-specific.
HPV16 genetic variation provides evidence of positive natural selection driven by HLA class I
HPV16 genomes contain 56 codons putatively under positive selection, disproportionately overlapping immune epitopes and divergent lineage sites, suggesting some variants improve persistence by escaping HLA-related immune recognition.
Natural genetic variation drives microbiome selection in the Caenorhabditis elegans gut
Natural genetic variation among wild C. elegans strains drives assembly of distinct gut microbiomes, with host insulin signaling mediating enrichment for Ochrobactrum.
Demography and Natural Selection Have Shaped Genetic Variation in the Widely Distributed Conifer Norway Spruce (Picea abies)
In Norway spruce, both purifying and positive selection can contribute to loss of genetic diversity by reducing variation.
Evidence that natural selection maintains genetic variation for sleep in Drosophila melanogaster
Latitudinal differences in Drosophila sleep behavior and associated gene-expression changes suggest natural selection maintains genetic variation for circadian behaviors.