aiAtlas: High Fidelity Cell Simulations of Genetic Perturbations in Rare Diseases and Cancers

Abstract

BackgroundLinking genetic perturbations to cellular phenotypes remains a central challenge in translational biology. Experimental iPSC and organoid models are powerful but constrained by scalability, variability, and difficulty modeling rare or polygenic states. MethodsWe developed aiAtlas v1.2, a high-fidelity simulation platform that integrates Large Concept Model (LCM) logic with aiPSC-derived modeling. We evaluated 136 virtual cell lines spanning wild-type, single-mutation, multiple-mutation, human tumor-derived, and gene-fusion cohorts. Twenty-five features covering DNA damage/repair, replication stress, epigenetic remodeling, pluripotency, and stress responses were quantified. Statistical analysis used the Mann-Whitney U test with Bonferroni correction, Hodges-Lehmann estimators (HLE) for median differences, and Cliffs delta effect sizes with bootstrap 95% confidence intervals. Robustness measures included early stopping, bagging, and 5-fold cross-validation. ResultsaiAtlas v1.2 reliably separated wild-type and mutant cohorts, revealing consistent disruptions in DNA damage accumulation, replication stress, epigenetic dysfunction, and loss of pluripotency, while identifying stable features (e.g., core nucleotide-excision repair processes and selected apoptosis measures). Subgroup analyses showed shared systemic effects and context-specific vulnerabilities: single mutations frequently produced measurable divergence; multiple mutations amplified instability; tumor-derived and gene-fusion lines yielded distinct but partially overlapping phenotypes. Large effect sizes (Cliffs {delta}) with narrow bootstrap CIs supported reproducibility across cohorts. Conclusions/ImpactaiAtlas v1.2 provides a robust virtual subject framework that uses aiCRISPR-Like (aiCRISPRL) virtual gene editing system that complements wet-lab CRISPR models by scaling to diverse genomic contexts and highlighting both disruption and stability. The platform can guide therapeutic prioritization, gene-editing strategy design, and regulatory innovation consistent with the FDA Modernization Act 2.0, accelerating therapy development in rare diseases and cancer. Significance StatementaiAtlas introduces a scalable, reliable simulation framework that integrates advanced large concept model (LCM) logic with iPSC-derived cellular modeling. aiAtlas overcomes major limitations of experimental systems by capturing both broad and subgroup-specific phenotypic divergence across single mutations, multiple mutations, tumor-derived cell lines, and gene fusions. This reliable platform establishes a new opportunity for rare diseases and cancer modeling, offering reproducible insights that can accelerate discovery and translational applications where traditional wet-lab approaches are impractical. Furthermore, in situations where the target mutational profile has been defined but no cellular models yet exist, aiAtlas can quickly generate custom virtual cell lines that accurately reproduce the corresponding genomic and phenotypic features.