Preclinical research relies extensively on animal experimentation despite well-documented limitations in translatability, cost, ethical burden, and scalability. In silico approaches have emerged as a promising complement to traditional experimental pipelines, particularly in alignment with the principles of Replacement, Reduction, and Refinement (3R). We present DATS Ultra (Digital Animal Twins), an in silico simulation framework designed to model population-level physiological responses of virtual animal cohorts to experimental interventions. The platform introduces a Dynamic Inference Engine for novel compound analysis and a 10-Point Organ Audit for systemic toxicity profiling. The platform enables parameterized simulations of substances, pathologies, environmental conditions, and lifestyle modifiers, producing longitudinal, normalized outputs relative to in silico controls. This work does not claim mechanistic causality or clinical predictivity. Instead, it establishes a conceptual and computational foundation for hypothesis generation, protocol prioritization, and experimental filtering prior to in vivo testing. By formalizing biological assumptions, parameter spaces, and system-level abstractions, this paper demonstrates the feasibility and scientific rationale of digital animal twins as a decision-support tool in preclinical research.