Older Pfam domain cohorts show less variance within them. The rate of decline matches that expected from simulations of inferred #DifferentialLoss rates. First time in my career that I was happy to get a small effect size! 8/11

We usually think of #DirectionalEvolution via descent with modification (top), but it's hard to imagine it causing that slow a trend. We hypothesized that #DifferentialLoss might be slow enough to explain why the trend kept going for billions of years without yet saturating 3/11