RE: Explains the continued and consistent contracts16 Aug 2025 12:03
From the same report,and PYC do 3 of these 4 recommendations
In Silico Tools and Computational Modeling In silico approaches are another pillar of NAMs. Computational modeling, artificial intelligence (AI), and machine learning (ML) can leverage existing data to predict safety, immunogenicity, and pharmacokinetics, reducing the need for new animal experiments. Key in silico tools include:• Physiologically-Based Pharmacokinetic (PBPK) Modeling: PBPK models are mathematical simulations of drug ADME (Absorption, Distribution, Metabolism, Excretion) using species-specific physiology. They have become integral in small-molecule drug development and are increasingly applied to biologics. FDA may review PBPK simulations to inform first-in-human dosing and to justify waiving animal studies that would normally serve that purpose. As PBPK models are refined, they can also predict how differences between patients (e.g. body weight, disease state) might affect a drug’s pharmacokinetics, further enhancing safety margins.• ML and AI Predictive Models: Machine learning algorithms can be trained on drug sequence features, structural motifs, and known clinical outcomes. Recently developed ML models analyze the amino acid sequence of an antibody’s variable region to predict whether the mAb is likely to have high or low immunogenicity (11). Such tools can flag problematic sequences early guiding engineering to “de-risk” the product before it ever enters an animal or human. Machine learning models are also being explored to predict toxicities (like acute systemic toxicity, off-target binding, or cytokine release potential) by learning patterns from molecules that caused certain adverse events (12).3 Roadmap to Reducing Animal Testing in Preclinical Safety Studies• Quantitative Systems Pharmacology (QSP) and Modeling of Biological Pathways: QSP models combine computational biology and pharmacology, simulating how a drug interacts with complex human biological networks. For example, a QSP model of an autoimmune disease could simulate how an antibody modulates inflammatory pathways, helping to predict efficacious dose ranges and potential toxic outcomes (such as over-suppression of the immune system). These models could reduce reliance on animal disease models by providing a virtual human on which to test “what-if” scenarios.• Bioinformatics and In silico Off-target Screening: Using databases of human proteins and AI, one could screen a product’s sequence for any unintended targets (such as cross-reactivity to human tissues). In silico tools can analyze whether the drug might bind to similar epitopes in the human proteome, highlighting potential safety concerns that would traditionally be checked via animal tissue cross-reactivity studies or broad receptor binding panels.
Sign In
Follow the stocks
