AI platform LUMI-lab sets record for gene-editing efficiency

Autonomous research platform LUMI-lab has achieved a gene editing efficiency rate of 20.3% in lung cells — a result that opens a new chapter in the application of artificial intelligence to biomedicine. This is not merely another iteration in the development of experimental tools: for the first time, a system based on foundational AI models has independently, without human intervention, discovered an mRNA delivery mechanism that has eluded researchers for many years.

Gene engineering has long suffered from one fundamental problem: between what scientists want to do with a cell and how to actually deliver the molecular 'cargo' into it lies a vast gap of trial and error. mRNA-based therapy — the very technology that underlies COVID-19 vaccines — requires precise delivery systems that depend on countless variables: the composition of lipid nanoparticles, their size, pH of the medium, specificity of target cells. Lungs as a target for gene editing are particularly complex: their cells exhibit high selectivity against foreign molecules, and an incorrectly chosen vector simply will not penetrate cell barriers. This is where the traditional scientific process — gradual, iterative, costly — shows its limits.

LUMI-lab is built on principles that the technology community calls 'autonomous research management.' The platform combines foundational AI models capable of processing and synthesizing scientific literature with robotic experimental setups. The system does not wait for instructions from researchers at each step: it formulates hypotheses, plans experiments, interprets results, and adjusts the next cycle of tests independently. This approach is fundamentally different from tools that merely accelerate the execution of human-defined tasks. LUMI-lab searches on its own — and finds where human intuition no longer copes with the scale of the search space.

The achieved 20.3% gene editing efficiency in lung cells is an indicator that has fundamental clinical significance in the context of gene therapy. To treat hereditary respiratory diseases such as cystic fibrosis, it is not enough to deliver a molecule to the organism — it must actually alter the function of a sufficient number of cells. Previous methods rarely exceeded a few percent under conditions approximating real tissue. The mRNA delivery mechanism discovered by LUMI-lab does not merely improve a number in a laboratory report: it establishes a new baseline for developing therapies that so far have remained at the level of promising concepts.

The significance of this achievement extends far beyond its specific application. Biomedical research has traditionally relied on a model in which the scientist is the central subject of each discovery. Platforms like LUMI-lab offer a different model: AI takes on the routine of experimental search, freeing humans for formulating more complex problems and interpreting results. This is not the displacement of researchers, but a redefinition of their role. For industry, this means reducing the time from hypothesis to clinically significant result — a cycle that today takes years could shrink to months. For pharmaceutical companies and biotech startups, this is direct savings on the most expensive part of development — the experimental phase.

At the same time, serious questions remain for this field. The reproducibility of results obtained by autonomous systems remains a subject of scientific debate: when an algorithm itself formulates the experimental protocol, independent verification requires additional effort. Regulatory bodies in medicine have not yet established clear frameworks for evaluating therapies developed with autonomous AI. This does not make LUMI-lab's achievement less significant — but it points out that the path from a laboratory record to approved treatment remains complex.

Nevertheless, the moment when an AI system independently discovered the 'code' for delivering mRNA to lung cells will enter the history of biomedicine as the point at which autonomous research platforms ceased to be a prospect and became a reality. The next question is no longer whether AI is capable of making discoveries — but how quickly the scientific community and regulators are ready to build an infrastructure of trust around these discoveries.