What happened
A research team at Fudan University Shanghai Cancer Center has developed a PARP-1-targeting PET radiotracer, [68Ga]Ga-DOTA-FZPF, using an AI-assisted design workflow. The process involved generative chemistry, computational screening, and preclinical validation in mice, culminating in an exploratory clinical study involving four patients with breast or ovarian cancer. The study, published in Materials Today Bio, marks the first reported instance of an AI-designed radiotracer undergoing human clinical testing.
The research team, led by Mengjing Ji, Xiangwei Wang, and Shaoli Song, utilized a multi-stage AI workflow to design the tracer. Starting with the PARP inhibitor fuzuloparib, they employed a SMILES-based conditional chemical language model to generate candidate analogs. These were then filtered using synthetic complexity models and convolutional -guided docking to ensure binding affinity to the PARP-1 protein.
Preclinical validation involved micro-PET/CT imaging in tumor-bearing mice, where [68Ga]Ga-DOTA-FZPF demonstrated high specificity for PARP-1 and significantly lower accumulation in the liver, spleen, and kidneys compared to existing tracers. The tracer also successfully functioned as a pharmacodynamic readout, showing reduced uptake in tumors treated with fuzuloparib.
The exploratory clinical study enrolled three breast cancer patients and one ovarian cancer patient. The tracer was found to be safe, with no adverse events reported. It showed clear uptake in breast lesions, though the researchers noted that further optimization is required to improve sensitivity for hepatic metastases.
Source details: bioengineer.org ↗
Why it matters
This development demonstrates that AI pipelines can effectively integrate generative chemistry, docking simulations, and synthetic complexity scoring to accelerate the discovery of radiopharmaceuticals. By automating the optimization of binding affinity and pharmacokinetics, this approach could significantly shorten the development cycle for molecular imaging agents. The probe specifically addresses a limitation in current PARP-1 imaging—high background accumulation in abdominal organs—by utilizing a highly hydrophilic design that improves imaging clarity in critical areas. This milestone provides a reproducible template for using AI to bridge the gap between molecular target identification and clinical application in oncology.
Current PARP-1 imaging agents often suffer from high background uptake in the liver and intestines, which obscures metastatic lesions in those regions. The AI-designed [68Ga]Ga-DOTA-FZPF addresses this by leveraging high hydrophilicity to reduce non-specific organ accumulation.
The integration of AI into the radiopharmaceutical design process represents a shift from traditional trial-and-error methods. By using computational tools to predict synthetic difficulty and binding stability, the team was able to move from a computer-generated structure to a human scan, establishing a new paradigm for rapid drug development in nuclear medicine.
Interactive Mechanism: How It Actually Works
Explore the underlying technology behind this development interactively.
crm_get_transaction(id='4092').Which component of an AI application is the machine-learning model itself?
What to watch next
Future research will focus on optimizing the tracer's tumor-to-liver ratio and conducting larger clinical trials to validate its sensitivity for detecting metastatic lesions. While the initial human study confirmed the safety and feasibility of the AI-designed probe, the authors noted modest tumor uptake in the small cohort, suggesting that further refinement of the molecular structure is necessary to improve diagnostic performance. The success of this workflow suggests that similar AI-driven pipelines could be applied to develop tracers for other biological targets, potentially expanding the toolkit for oncology.
The researchers have identified the need for larger clinical cohorts to better understand the tracer's diagnostic accuracy across different cancer types and stages.
Future iterations of the AI design pipeline may focus on balancing the tracer's hydrophilicity with improved cell membrane penetration to enhance tumor uptake, which was described as modest in the initial human trial.