A new breed of artificial intelligence is shifting biology from reaction to prediction. Instead of waiting for drug resistance to emerge or populations to collapse, researchers are now training AI systems to forecast biological change by studying how traits evolved across millions of species over deep time. Astromech, an evolutionary biology AI company, just raised $20 million to scale this approach, bringing its total funding to $60 million and valuing the company at $3.8 billion.

Why Can't We Just Predict Biological Change Today?

Biology moves fast, but our ability to anticipate it lags behind. Researchers can describe genomes in extraordinary detail, yet they struggle to predict what comes next. Most computational biology models analyze organisms as they exist today, treating each species as a snapshot rather than a story. That snapshot approach misses the evolutionary context that reveals why certain traits emerge, persist, or fail.

The challenge runs deeper than raw computing power. Every genome carries a record of what changed, when it changed, and the tradeoffs that followed, but almost none of that history is readable at scale today. Astromech was built to close that gap by reconstructing not just ancestral proteins, but ancestral regulatory states, chromatin accessibility, and gene expression patterns across species and deep time.

"Biology runs the world and historically, we have only reacted to it. We can describe biology in extraordinary detail, yet we still struggle to anticipate what comes next. Astromech is building the AI system predicting how living systems will change and where they are most likely to break," said Ben Lamm, co-founder of Astromech.

Ben Lamm, Co-founder of Astromech

How Does Astromech's Predictive Model Actually Work?

Astromech's architecture ingests three layers of data: genomic (living and extinct genomes), evolutionary (deep-time ancestry and divergence), and functional (gene expression, traits, and environmental responses). From those inputs, the system produces three kinds of forecasts: where a genome or population is headed, where that biological system is most likely to break, and which regulatory circuits are driving the change.

Two computational engines sit underneath. The first is a suite of deep learning models that find patterns across species and systems, learning how genes are expressed and how organisms respond to their environment. The second works backward through evolutionary history, reconstructing how a system reached its current state, then runs the same mathematics forward to project where it goes next. Fused together, they form a single model system trained on the history of biological change across deep time.

The technical core extends ancestral state reconstruction beyond sequence alone. Rather than inferring only the ancestral protein at each node of a family tree, Astromech reconstructs ancestral regulatory state. It integrates that evidence across multiple data modalities into a Bayesian framework that returns calibrated confidence rather than point predictions. In internal benchmarks, the company developed a learned tree-inference method that reconstructed evolutionary trees approximately 100 times faster than conventional maximum-likelihood methods while maintaining comparable accuracy.

Steps to Leverage Predictive Genomic AI in Research and Medicine

• Assemble Comparative Genomic Data: Gather functional genomic data across multiple species, not just one reference genome. Astromech trains its models across species, supplementing public datasets with functional data generated in-house for species where coverage is limited, allowing models to examine regulatory changes between lineages rather than variation within a single species.
• Reconstruct Ancestral Regulatory States: Move beyond protein-coding sequences to map how genes are regulated across evolutionary time. For complex traits like longevity and disease resistance, important variation occurs outside protein-coding regions and affects how genes are regulated, so reconstructing ancestral regulatory state is essential to understand trait evolution.
• Validate Predictions Against Known Outcomes: Test the model's ability to recover trait-associated genes previously established in published research. In retrospective validation, Astromech's pipeline recovered known trait-associated genes while identifying additional candidates for further study, providing a foundation for prospective validation through future partner pilots.

What Real-World Problems Could This Solve?

Astromech's platform is designed to generate three types of biological insight: where a genome, pathogen, or population may be headed; where a biological system may be most vulnerable, including potential susceptibility to disease or drug resistance; and which regulatory mechanisms may be driving those changes. The company identified several potential applications across human health, biosecurity, agriculture, and conservation:

• Disease Surveillance: Flag susceptibility across species to different pathogens before they reach humans, enabling early intervention and public health planning.
• Drug Resistance Prediction: Predict drug resistance before it becomes a treatment failure, allowing clinicians to rotate therapies or adjust treatment strategies proactively.
• Disease Risk Mapping: Map disease risk and the drivers of healthspan, identifying which regulatory changes contribute to longevity and disease resilience.
• Population Vulnerability Assessment: Model herd vulnerability under disease and climate stress, supporting conservation and agricultural planning.

How Does This Fit Into the Broader Digital Biology Boom?

Astromech's funding round arrives as the global digital biology market is accelerating rapidly. The market is expected to grow from $17.35 billion in 2026 to $46.12 billion by 2033, registering a compound annual growth rate of 15 percent. Artificial intelligence and machine learning is projected to hold 31.6 percent of the global digital biology market share in 2026, driven by their ability to analyze complex biological data sources, detect patterns, and generate predictive insights.

Government investment is fueling this expansion. In July 2026, the U.S. National Science Foundation announced $83 million in awards to integrate scientific data with computing and AI resources, supporting large-scale, data-intensive research and AI-driven discovery. In January 2026, the European Medicines Agency and U.S. Food and Drug Administration established 10 common principles for good AI practice in medicine development, covering AI use across research, clinical development, manufacturing, and safety monitoring.

North America maintains dominance in the digital biology market with an expected share of 40.8 percent in 2026, bolstered by its established biomedical research infrastructure, extensive genomic datasets, and advanced computational capabilities. The U.S. National Institutes of Health All of Us Research Program provides researchers with data from more than 747,000 participants, including over 535,000 whole-genome sequences linked to nearly 482,000 electronic health records.

What Makes Astromech's Approach Different From Other Genomics AI?

Most computational biology models treat genomes as static objects to be analyzed in isolation. Astromech treats genomes as historical documents, reading the evolutionary narrative encoded within them. By studying how biological traits developed over time, the company can identify traits including longevity, cancer resistance, and tolerance to environmental stress that have evolved independently across many species, creating naturally tested examples of biological resilience.

The breadth of training data matters enormously. Models that predict regulatory function from sequence are typically trained on one or two reference genomes, where most functional genomic data has historically been concentrated. Astromech trains its models across species, supplementing public datasets with functional data generated in-house for species where coverage is limited. This broader comparative foundation allows the models to examine regulatory changes between lineages, rather than variation within a single species.

Astromech was gestated at Colossal Biosciences and has access to the genomic data resources Colossal has spent five years assembling: a broad genome bank of extinct and living species, tooling battle-tested on massive biological datasets, ancient-DNA capability that reads old genomes and compares them to living ones to see exactly what changed and when, and a scientific team trained across evolutionary systems and computational biology.

The $20 million funding round, led by biotech investor Bob Nelsen and including participation from Peak 6, NeoGenesis Capital, Builders VC, and CAZ Investments, will allow Astromech to expand its research team, increase the number of species represented in its functional genomic datasets, and scale the comparative genomic infrastructure used to train its models. As genomic AI moves from describing biology to predicting it, Astromech represents a fundamental shift in how researchers approach some of biology's oldest questions.