Astromech raises $20M to forecast how living systems will evolve

Astromech was spun out of Colossal Biosciences Inc., a genetic engineering firm working on de-extinction of animals including the woolly mammoth and Tasmanian tiger. That heritage shaped the company's core asset: 3.8 billion years of evolutionary history as a primary training signal. The dual-engine architecture—one using deep learning to spot patterns across species, the other to reverse-engineer and forward-project biological systems—reflects a conviction that evolution itself is the richest source of data for predicting biological change.

Co-founder Church's emphasis on ancestral regulatory states rather than just proteins addresses a technical limitation in existing comparative genomics: most genetic variations affecting complex traits like morphology and longevity are regulatory, not coding. By reconstructing ancestral regulatory states, Astromech claims to recover explanatory power that protein-only comparisons miss. The focus on longevity—studying species like the Asian elephant's cancer-suppression mechanism and the bowhead whale's extreme lifespan—is framed as a near-term validation target, not a near-term product. Lamm's weather-forecasting analogy positions the startup as solving an inference problem (predict future states from historical data and current conditions) rather than a protein-design problem, which may appeal to investors and partners seeking forecasting rather than engineering.