Astromech secured $20 million at a valuation of $3.8 billion to predict how biology fails.

      Astromech has secured $20 million at a valuation of $3.8 billion to develop AI that forecasts changes in living systems. The company claims its models are informed by 3.8 billion years of biological history. The alignment of the valuation and the training duration is noted, but no one has commented on whether it is coincidental. The funding round was led by Bob Nelsen and included Peak 6, NeoGenesis Capital, Builders VC, and CAZ Investments, according to GamesBeat. The total funding now amounts to $60 million.

      Astromech was founded by Ben Lamm and George Church, the duo behind Colossal Biosciences, which aims to revive the woolly mammoth. Conversely, Astromech focuses on predicting biological developments rather than reconstructing past events. Based in Dallas, the startup cultivated its foundation within Colossal and utilized its data resources, which encompass a genome bank of extinct and extant species, tools for handling extensive biological datasets, and capabilities for comparing ancient DNA with current genomes to identify changes over time.

      Lamm describes the effort as a forecasting initiative. “We are developing an algorithmic prediction solution,” he explained to Inc. “Consider it like weather forecasting, where humanity can predict outcomes due to specific technologies and datasets. We aim to apply the same principles to biology using evolutionary data.” The platform assimilates three types of inputs: genomic, evolutionary, and functional, and generates three outputs: predictions for genome or population trajectories, points of potential breakdown, and the regulatory circuits influencing these changes. “Biology governs the world, and historically, we have only responded to it,” Lamm stated in the announcement.

      The technical approach involves ancestral state reconstruction beyond mere sequence analysis. Instead of inferring only the ancestral protein at each node on a phylogenetic tree, Astromech reconstructs the ancestral regulatory state—considering chromatin accessibility, gene expression, and functional annotations. This process employs a Bayesian framework, yielding calibrated confidence levels rather than singular predictions. Church elaborated on the significance, noting that most variations pertinent to complex traits, such as morphology and longevity, stem from regulatory rather than coding elements, making the reconstruction of ancestral regulatory states vital for explanatory power. He emphasized that achieving this requires functional data across numerous species and affordable AI reconstruction for genome-wide applications—conditions not feasible a decade ago.

      Currently, the company is in what it describes as an extensive research and development phase. Its first public demonstration illustrates 46 genes associated with longevity based on a time-calibrated tree of life. It contrasts species that have tackled similar challenges in diverse ways, such as Asian elephants that resist cancer despite their size, bowhead whales that live for over two centuries, Brandt’s bats that, weighing only a few grams, exceed forty years of life, and Tasmanian devils, which are vulnerable to a transmissible cancer. Astromech claims its internal benchmark for tree inference operates approximately 100 times faster than traditional maximum likelihood methods while maintaining similar accuracy. In retrospective validation, the pipeline successfully identified previously established trait-associated genes from published studies and highlighted additional candidates.

      This represents the entirety of the public record. Validation of forecasts, meant to determine their accuracy, will be conducted in the future through partnerships in health and biosecurity. This differentiation is crucial; recovering genes identified in established research indicates the pipeline is functioning on known outcomes, yet does not validate the model's ability to predict new findings in advance.

      The numbers are striking: $60 million raised, with a valuation of $3.8 billion, one instance of retrospective validation, and a mapping of 46 genes. This equates to approximately sixty-three times the funds raised, even though the platform has no market product. Investors have previously supported this founding team through extended timelines. Tech Funding News noted that Colossal Biosciences invested five years and hundreds of millions before yielding a de-extinct organism and continued to attract investment regardless. The valuation is consistent with similar contexts; TNW reported in July that Chai Discovery also raised $400 million at the same $3.8 billion valuation, albeit with a different issue and significantly higher capital.

      Other entities in this sector include Insitro, which has amassed over $700 million for machine-learning drug discovery, and Alphabet’s Isomorphic Labs. Astromech distinguishes itself by positioning upstream; it does not design drugs but aims to predict biological failures before specific targets are chosen. A Cambridge startup similarly raised $20 million in July to develop a foundational model for the chest, intending to expedite drug trials. Additionally, Google DeepMind ventured into overlapping areas in July with an AI biosecurity initiative. The foundational raw materials continue to improve as well, with the NIH unveiling the largest genomics and health database in July. Investment activity is thriving in related spaces; Novartis has committed up to $1.5 billion to the UK biotech Myricx this past July.

      Astromech outlines six applications for its platform: identifying species vulnerable to