Astromech secured $20 million at a valuation of $3.8 billion to predict how biological systems fail.

      Astromech has secured $20 million at a valuation of $3.8 billion to develop AI that anticipates changes in living systems. The company claims its models are informed by 3.8 billion years of biological evolution. Interestingly, the valuation and the training duration share the same figure, but it's unclear if this is merely coincidental.

      Bob Nelsen led this funding round, which included contributions from Peak 6, NeoGenesis Capital, Builders VC, and CAZ Investments, as reported by GamesBeat. The total funding for Astromech has now reached $60 million.

      The founding team

      Astromech was founded by Ben Lamm and George Church, who are also behind Colossal Biosciences. Colossal aims to resurrect the woolly mammoth, while Astromech takes a different approach, focusing on predicting future biological developments instead of reconstructing past ones.

      The Dallas startup originated within Colossal and has access to its data, which includes a genomic library of extinct and extant species, tools designed for substantial biological datasets, and capabilities for comparing ancient DNA with contemporary genomes to track changes over time.

      Objectives of the platform

      Ben Lamm describes the company’s goal as making predictions. He stated to Inc., “We are creating an algorithmic prediction solution, similar to weather forecasting, which is possible due to specific technology and data sets. We aim to do the same for biology using evolutionary information.”

      The platform integrates three types of inputs—genomic, evolutionary, and functional—and generates three outputs: predictions about the trajectory of a genome or population, points of likely failure, and the regulatory mechanisms driving these changes. Lamm emphasized, “Biology governs the world, and historically, our response has only been reactive.”

      Technical approach

      Astromech employs a method called ancestral state reconstruction, which goes beyond just sequencing. Instead of merely deducing the ancestral protein at various points on a phylogenetic tree, it reconstructs the ancestral regulatory state, including chromatin accessibility, gene expression, and functional annotation, using a Bayesian framework that provides calibrated confidence, rather than singular predictions.

      Church highlighted the importance of this approach, noting that significant variations affecting complex traits such as morphology and longevity are more often regulatory than coding. Thus, reconstructing the ancestral regulatory state is crucial. He added that obtaining functional data from diverse species and having affordable AI reconstruction for genome-wide analysis were not feasible a decade ago.

      Current status

      Astromech currently identifies itself as being in a phase of deep research and development. Its first public demonstration highlighted 46 genes associated with longevity mapped across a time-calibrated tree of life, comparing species that have developed different solutions to similar challenges: Asian elephants with cancer resistance relative to their size, bowhead whales living for over two centuries, Brandt’s bats weighing only a few grams but living beyond forty years, and Tasmanian devils, which are susceptible to transmissible cancer.

      The company claims that its tree-inference methodology runs approximately 100 times faster than traditional maximum-likelihood methods while maintaining comparable accuracy. In terms of retrospective validation, its pipeline was able to identify genes associated with traits already documented in existing research and suggested additional candidates.

      This constitutes all publicly available information. Prospective validation—testing whether its forecasts hold—will occur later through collaborations in healthcare and biosecurity.

      Financial overview

      With $60 million raised and a valuation of $3.8 billion, Astromech has established a value that is about sixty-three times its capital raised, all while lacking a product on the market. Investors have previously supported this founding team during lengthy waits: Colossal, for example, invested five years and hundreds of millions without successfully creating a de-extinct animal yet continued to attract funding.

      The valuation is consistent with recent financial activities; as TNW reported in July, Chai Discovery also raised $400 million at a similar $3.8 billion valuation, addressing a different issue and securing twenty times the funding.

      Competitors in the field

      Tech Funding News mentions companies like Insitro, which has raised over $700 million for machine-learning drug discovery, and Alphabet’s Isomorphic Labs. Astromech differentiates itself by working upstream of these entities, focusing on identifying biological failures before any specific targets for drug development are chosen.

      There is extensive activity downstream as well, with Novartis recently spending up to $1.5 billion to acquire the UK biotech Myricx. Additionally, in July, a Cambridge startup secured the same $20 million to create a foundational model for the chest, aimed at accelerating drug trials, while Google DeepMind ventured into nearby territory with an AI biosecurity initiative. The availability of data is also advancing, as the NIH unveiled its largest genomics and health database in July.

      Intended applications of the platform

      Astromech outlines six potential applications: identifying species at risk of exposure to pathogens before reaching humans; predicting drug resistance ahead of treatment failures; assessing disease risks and healthspan; modeling