Much of the beauty — and challenge —of biology lies in its complexity. That’s especially true in the microbial world, where hundreds or thousands of different bacterial species may coexist in a patch of soil or in a section of the human gut. Each species has its own way of life, but each also interacts with the others to shape the ecosystem.

But is there hope for some order in that complexity? A new paper in Science co-authored by Mikhail Tikhonov, an associate professor of physics, has pierced through the apparent chaos to find surprising levels of predictability in microbial systems.

“Ecologists and physicists have long had an intuition that the extremely large number of species in these ecosystems might paradoxically make them easier to understand,” Tikhonov said. “Here, we’ve shown that indeed, under some conditions, increasing the complexity of a system makes simple models increasingly predictive. But importantly, predictability is not automatic, and it’s not guaranteed by complexity alone.”

The paper — co-authored by physics graduate student Lucas Graham and former graduate student Jacob Moran, PhD ’23 — avoids the easy observations that have long clouded the picture of complex ecosystems. “One obvious explanation is that when you add a lot of different species, the quirks average out,” Tikhonov said. “That’s a very reasonable assumption, but it’s insufficient. It doesn’t deliver actual predictions.”

Tikhonov, Graham, and Moran went beyond that facile explanation to reframe the question. Instead of asking if complex ecosystems can be predictable, they tackled a deeper issue: How much better do predictions become after adding new information about the composition of the community? “Crucially, we can formally quantify the amount of information needed to improve predictions,” Tikhonov said. “This means that predictability isn’t just a vague concept. It’s something that can be measured and tested.”

Tikhonov and his co-authors distinguish three different “flavors” of simplicity that can arise in an ecosystem. As they explain, the first is a trivial consequence of statistics, the second is an expected outcome of species relatedness (closely related species behave similarly), and the third reflects genuinely interesting biological structure — the kind that can support meaningful predictions. “Our main contribution is isolating that third case,” Tikhonov said. “That’s the one that actually tells you something about how the system works.”

The value of making these distinctions, Tikhonov said, is not in splitting hairs, but in understanding what kinds of simplicity are informative and which are not. He compared this to how scientists learned to be more precise about the idea of randomness. “Many very different things can be colloquially described as random,” he said. “But over time, scientists learned to make careful distinctions: Some apparent randomness is the noise of imperfect measurements, some systems are genuinely stochastic, and some exhibit chaotic dynamics even with no randomness at all. Those distinctions matter enormously, because they tell you what kind of explanation to look for, and what kind of prediction is even possible.”

In the same way, Tikhonov said, grouping all apparent regularities in complex ecosystems under a single notion of “simplicity” can obscure what is actually going on. Some apparent simplicity is inevitable and uninformative, while other kinds point to underlying biological constraints.

Using data from lab-assembled communities of bacteria drawn from soil and the human gut, Tikhonov and his team were able to test these distinctions directly. As expected, the most trivial form of simplicity — simple statistical averaging — was always present. But strikingly, so was the more interesting kind: As diversity increased in their lab-assembled communities, coarse information about community composition became increasingly predictive of biological outcomes, such as fermentation products.

Better predicting the activity of soil or gut bacteria could someday lead to new advances in agriculture or healthcare, but those are possibilities for another time, Tikhonov said. The more immediate contribution of the work is conceptual: clarifying when diversity can make ecosystems more predictable. “To really understand these systems,” he said, “we need our theoretical language for simplicity and predictability in complex ecosystems to be more precise.”

Header image: Photo by Edward Jenner/Pexels