Welcome back.

For most of medical history, we’ve studied human disease by studying humans. Duh, of course, why wouldn’t we? But it turns out that when medicine starts looking at the DNA of other species, we can unlock so much more knowledge about ourselves.

Tech and climate journalist Tasmin Lockwood returns this week to explore how cross-species biology, supercharged by AI, is beginning to rewrite what we thought we knew, and what the DNA of animals like elephants can tell us about our own health.

Enjoy!

— Willow

Trekking through the Bolivian Amazon in 2024, I was acutely aware of nature’s pharmacy and kitchen that lay around me.

Close to our entry point, deforested land was now a small banana plantation. I was reminded of the fact that humans share around 50% of their genes with the yellow fruit — a fun fact I giggled at as a teen. Deeper into the rainforest, my guide carved the bark of a tree and passed around his findings. Ah, cinnamon, a tasty natural anti-inflammatory. 

The health of humans and the natural world are typically discussed in isolation, but they are innately interlinked.

Studying the evolutionary biology of non-human species offers scientists a chance to better understand the common building blocks of life. It reveals things about human biology that human-only science misses, including which mutations can be harmful. 

But there are millions of known species on Earth, presenting a daunting task. Now, AI is unlocking the ability to use evolutionary data for medicine with the scale and nuance it needs to spot mutations and rare diseases. 

Evolutionary medicine is a crucial line of inquiry in the age of the climate crisis. Biology from the animal kingdom can flag human genetic mutations likely to cause disease, even in the rarest of cases. 

It also unlocks better prediction and prevention of pathogens, offering a possible line of defense against the spread of disease. Finally, by assessing how other species react to conditions, scientists can establish new potential treatments for humans.

For example, elephants are estimated to have up to four quadrillion cells, compared with a human’s 40 trillion. With such vast body mass, scientists have historically been puzzled by why they rarely develop cancer, which occurs as damaged cells accumulate. 

The general understanding is that elephants have at least 20 copies of a specific protein, p53, which repairs damaged cells — including cancerous mutations. Humans only have one. It presents an opportunity for new therapies that could be applied to humans, though these are still in the early days of exploration. Meanwhile, the TP53 gene, which encodes the p53 protein, is one of the most frequently mutated genes found in cancer patients. 

By tracking which positions in a protein’s structure have stayed consistent across species over time, scientists can identify those likely to play a crucial role in the protein’s function. If a mutation appears in a constrained position, it carries a higher potential of harm as it can alter a core function of the protein. Positions that biology has allowed to change are likely to be low risk.

One of the best ways to map out those constrained positions is to look at the biology of all species and identify where they converge and diverge. For instance, if a position is the same across mammals, birds, and fish, scientists can assume it’s important. 

Google DeepMind’s aptly named AlphaMissense model — missense being the name of the mutation that changes a protein’s structure or function — was one of the first mainstream attempts to do this using artificial intelligence.

It is built on top of its AlphaFold system, which caused shockwaves when it predicted structures for most of the proteins known to science, though it was primarily focused on human and primate mutations. 

A newer AI model, known as popEVE, outperforms AlphaMissense and other competitors at identifying when humans may carry gene mutations. It was developed by researchers out of the Centre for Genomic Regulation in Barcelona and Harvard Medical School and trained on thousands of animal species, according to a late 2025 paper in Nature. 

In one batch of data of patients with severe developmental disorders, it found 4.4-times more candidate novel genetic disorders than previously identified, a reminder of how many genetic mutations remain undiagnosed or unidentified.

Even alone, it makes a compelling case for using cross-species data to boost science’s understanding of genetic variants in spaces it was previously flying blind. 

The caveat, as noted by the authors of the study, is that AI is incredibly energy-intensive to run. It also uses as much water as the bottled water industry each year. While new techniques may be leveraging evolutionary science to diagnose rare conditions, we need to make sure that they don’t destroy the planet — and the very biodiversity it relies on — in tandem. 

Thanks to its ability to crunch vast amounts of evolutionary data, AI can also help to spot evolutionary tricks that could be used as medicine, thus supercharging both diagnosis and treatment.

Evolution has essentially run millions of experiments for millions of years; the hope is that by assessing protein divergence in species with useful traits, such as the elephant’s natural resistance to cancer, findings could be translated into human therapies. 

While some progress has been made since, older research suggests improper folding of the p53 protein could lead to ineffective treatment or even more harm when given to cancer patients. AlphaFold could be another unlock here as it can predict protein structures to a high accuracy. 

A significant amount of modern medicine already originates from the natural world. Together with indigenous knowledge, nature accounts for 40% of pharmaceuticals. Aspirin comes from willow bark; penicillin derives from mold; and the original sources of some chemotherapy drugs are found in nature.

This has given rise to the field of biodiscovery, which is the process of exploring nature to find new systems, materials, or processes that could have applications for humans, such as medicines or biomaterials.

British company Basecamp Research, for instance, is building a knowledge graph of nature to map the world’s biodiversity while discovering and designing  useful proteins for researchers and industry.

Most of the world’s biodiversity remains uncharted; global scientists name around 2,500 new-to-science plants every year.

Nature loss threatens potential medicines and important biological information before we even understand what we’re losing. That means we don’t have a full, holistic view of how species relate to one another, and understanding this is core to ensuring ecosystems are healthy, which in turn helps to keep disease in check — for humans too.

The discussions linking planetary health, biodiversity integrity, and human health are at the early stages, but full of potential. AI could also be used to monitor how biodiversity is evolving, changing, and eroding, helping to anticipate what ecosystem changes may come as a result. 

For instance, as the Earth warms and their predators decline, mosquito populations grow and expand into new areas. That, in turn, exposes more people to deadly diseases. Diseases from mosquitos, flies, and ticks are responsible for 17 % of the world’s infectious diseases, which is only worsening as their density increases. 

Evolutionary science and AI could shed light on these trends, and aid mitigation and adaptation. Research into an AI project to identify mosquito species in real-time from geolocated images found that it helped identify disease-carrying species, meaning action could be taken to counteract a potential health crisis. 

The current lack of cross-species biological comparison reveals structural issues with the way science is practiced. 

Humans see themselves as separate from nature, which is visible in the way that we treat other species, be it through land use, industrialized farming, or the overexploitation of resources.

How that feeds back into human health is easily overlooked; declining biodiversity exacerbates climate instability and ecosystem collapse, which directly threatens medicine, food, and water security. 

Even without AI, science often exists in silos and cross-departmental researchers rarely exchange notes. If they did, profound breakthroughs could be made. 

Research into porous coral reefs, for instance, has borrowed imaging tools from human osteoporosis diagnosis. It shows how similar characteristics can be found in unexpected places and emphasizes how lessons can also be learned in reverse, by looking at human health and applying it to nature. 

The human body also isn’t siloed; at a physiological level, a small issue in one part of the body can manifest as a much larger issue in another.

Proteins are made up of chains of amino acids; they work closely with DNA, acting as the structural and functional form of data that is encoded into DNA. While mutations occur at this level, changes are seen in amino acids. Life at this scale is invisible but reveals the profound interconnectedness of even the smallest of biological nuts and bolts. 

Knowledge graphs, like the one built by Basecamp Research, embody the spirit of nature in this way by identifying connections between big and small ecosystem players.

Humans used to live like this too. I was reminded of exactly that when visiting the south of Spain with friends in April: community-first spaces, children playing together harmlessly, parents laughing with one another and enjoying the city’s golden hour as the sun set. Many countries — cities, even — no longer live with an ecosystem approach, instead favoring individualistic systems. 

Perhaps this is also something that can be learned from elephants. Known for their intelligence and long lives, they are herd animals that form deep personal bonds with relatives and different family groups. 

Humanity’s inability to think in systems is one of its greatest weaknesses today, and this translates into how medicine is practiced. An ecologically inspired mindset is one that doesn’t try to optimize or sustain a system as it is, rather it is designed to be alive, resilient, and able to regenerate itself. 

Applied to medicine and evolutionary biology, facilitated by AI, this could simply mean asking the model different questions: what interventions could be taken to improve many parts of the system, rather than just one isolated mutation?

And what action can be taken or what other data could be considered to understand this risk better in the future?

Together, it could unlock meaningful preventative care, rather than reactive but often temporary fixes.