Background: Limitations of LNP Liver-Targeting Bias and Bottlenecks in Cell-Resolution Genetic Data for Autoimmune Disease Drug R&D

• Conventional, linear, and static analysis guidelines have a critical blind spot: liver-centric tropism, an inherent bias where LNPs are forced to target hepatocytes due to adsorption with the serum protein ApoE during intravenous administration. This hinders the delivery of genetic material to cells with low membrane permeability, such as those in tumor microenvironments, bone marrow, spleen, and macrophages, leading to cell lysis-induced structural collapse noise and preventing effective engraftment and maintenance of therapeutic concentrations. Furthermore, cell lysis noise and resistance feedback loops prevent computational control of the dynamic tensor of non-liver organs. Consequently, omics-based, multi-dimensional interaction matrices for in vivo cell reprogramming have remained static, failing to find a balance between target cell delivery and biocompatibility, and causing bottlenecks in metabolic and genetic data.

Discovery: Implementation of the RAPTOR Machine Learning Algorithm and Demonstration of Multi-Cell Scale Ionizable Lipid Variable Tensor Synchronization

• As published in Nature Biotechnology (doi:10.1038/s41587-026-03201-5), Liberate Bio's RAPTOR platform surpasses existing LNP exploration models. This platform employs a large-scale machine learning architecture to proactively design novel cationic ionizable lipid libraries in silico, and performs high-throughput pooled screening at the scale of animal models, including non-human primates, to perfectly synchronize independent variable tensors. This allows for fine-tuning of the free energy constant between lipids and the calculation of delivery rate constants based on differential equations, thereby computationally eliminating batch effects. As a result, the topological baseline variation curve of downstream transcriptome networks is elucidated, and high-resolution delivery efficiency and genetic integrity are demonstrated for previously inaccessible cells in non-liver target microenvironments.

In Vivo CAR-M Cell Therapy Pathway Modulation and Establishment of a Reversible Immune Homeostasis Precision Layering Model

• The RAPTOR LNP toolbox enables the creation of an immunological homeostasis control model based on multi-dimensional omics matrices, implementing individualized molecular phenotypes and precision stratification for patients. Specifically, it precisely controls the rate-limiting step constants of the In Vivo CAR-M (Chimeric Antigen Receptor Monocyte/Macrophage) cell therapy molecular pathway by up-clamping or down-clamping resistance pathways. This allows for the modulation of phenotypic modification factors of target immune cells, such as monocytes, even under the aberrant stress conditions of autoimmune diseases. As a result, a backbone architecture is established that dynamically maintains reversible immune homeostasis in vivo, maximizing personalized genome delivery therapy tailored to individual genetic gradients.

Prospects: Establishing a Standard for Programmable Computational Genomics and Implementing Next-Generation IND Digital Governance

• This computational genomics delivery technology heralds a reset of drug R&D governance, which has relied on post-hoc symptomatic treatment, into a programmable infrastructure based on multi-dimensional tensors. By linking genetic gradient correction coefficients to the high-throughput screening (HTS) stage of global multinational pharmaceutical and biotechnology pipelines, a computational moat is established that eliminates batch-to-batch variation in large-scale commercial production. This will meet companion diagnostic (CDx) standards, improve patient selection efficiency, and serve as a master asset that disruptively shortens the timeline for clinical trial (IND) approval and cGMP commercial launch by global regulatory agencies.