Automation ≠ Accelerated R&D: Ou Shisheng Helps Academician Qian Xuhong’s Team Achieve Five-Step Fully Continuous Synthesis of Zidovudine, Empowering Continuous Intelligent Pharmaceutical Manufacturing
The 4th Annual Conference on Microchemical Engineering Technology, organized by the Chemical Industry and Engineering Society of China, took place in Nanchang from October 16th to 19th, 2026. During the plenary session, Professor Qian Xuhong, a Member of the Chinese Academy of Engineering and affiliated with East China Normal University, presented on the five-step, fully continuous synthesis and purification of zidovudine, an anti-AIDS medication, using it as a key example. He thanked Ou Shisheng and showcased the intelligent continuous drug synthesis equipment that Ou Shisheng had specifically developed for his research team.
01 Introduction: Practical Requirements of AI4S for the Experimental Layer
Before detailing this case, it's important to explain Ou Shisheng’s understanding of AI4S. The demands AI4S places on the experimental layer are not simply about adding automated modules or increasing throughput. Automation focuses on "unattended operation," and throughput measures "sample count per unit time." Neither addresses a more fundamental question: are the reactions being run under their optimal kinetic and transport conditions? In traditional batch reactions, even with advanced automation, the primary result is just freeing up labor. Accelerating R&D by reducing manpower and enabling 24/7 experimentation offers limited benefits with a clear ceiling, failing to fundamentally shorten the R&D cycle.
The true key to efficient data generation lies in process knowledge at the reaction stage. This includes enhanced mass and heat transfer in microchannels, precise control of residence time, and wider temperature ranges under pressure. These elements lead to significant improvements in reaction rates and space-time yield compared to manual batch operations. The closed-loop AI4S model depends on the experimental layer's ability to generate data effectively. The "hypothesis–validation–iteration" data cycle can only spin sustainably if the rate of data production matches the model's learning rate. The significance of the five-step fully continuous zidovudine process lies in these substantial improvements in reaction engineering, not in overly complex hardware.
02 Inherent Limitations of Batch Processes in Classic Pharmaceutical Synthesis
Zidovudine (AZT), a well-known anti-HIV medication, has a synthetic pathway that begins with thymidine and involves five sequential reactions: hydroxyl protection, sulfonylation, intramolecular cyclization, azidation, and deprotection. This route has been carried out in batch reactors for many years. While the process is well-established, its inherent limitations have also become deeply ingrained. The five reaction steps are conducted separately; each step requires isolation, purification, and reintroduction of materials. Intermediates are moved between vessels, resulting in production cycles that can last for weeks.
The fourth step is especially crucial. The azidation reaction uses sodium azide at elevated temperatures. The prolonged presence of large quantities of material within batch reactors makes this stage the primary safety concern for the entire process. The challenge arises not from the reaction's feasibility, but from the safety restrictions encountered during scale-up. This is a longstanding issue with batch processes: achieving adequate capacity, safety, and consistent product quality simultaneously is difficult.
03 How the Five-Step Reactions Are Integrated into a Single Flow Sequence
The process workflow presented in Academician Qian Xuhong’s presentation offers an alternative solution: instead of being carried out in separate reactors, the five reaction steps are cascaded into one continuous flow line. Feed enters at one end and exits as the final reaction product at the other. No offline separation of intermediates is performed along the way; the crude product is obtained only after downstream workup.
Figure 1 | Five-Step Fully Continuous Zidovudine Synthesis: A series of five microreactors with integrated inline HPLC, back pressure regulators, and ultrasonic water bath, coupled downstream with modules for precipitation, filtration-washing-dissolution, distillation, and crystallization.
Three key engineering features are incorporated into this continuous flow line.
First pressure regulators, operating at 100 psi and 75 psi, keep the system under pressure. This allows several steps to run stably at temperatures of 120–180 °C, well above the atmospheric boiling point.
Second, the azidation reaction takes place within a reaction volume of just 1.1 mL. This dramatically reduces the on-site inventory of hazardous materials from kilograms in batch reactors down to the milliliter scale. Reagents are consumed as soon as they are produced, which is the practical meaning of inherent safety.
Third, online HPLC units are installed at several points along the line. This eliminates the need for offline sampling and analysis to determine reaction conversion; instead, conversion is monitored in real time as the stream progresses.
04 Post-Reaction Operations: Workup Integrated into the Full Continuous Sequence The five reaction steps constitute only the first half of this production line. The cyclized product first undergoes online solid-liquid separation and solvent exchange via the precipitation module and the integrated filtration-washing-dissolution module before entering the azidation stage. After deprotection, the stream passes sequentially through distillation, a second integrated module, secondary distillation and crystallization modules, and is finally discharged as crude zidovudine.
Integrating workup into the continuous chain carries significance comparable to the cascade of reaction steps. In batch processes, separation and purification usually account for most of the cycle time and labor demand. This stage also marks the dividing line between batch and continuous manufacturing. If only the reactions are operated continuously while workup remains batchwise, the production line will still be pulled back into a batch rhythm by offline intermediate separation, transfer and re-feeding. The integrity of this process lies in uninterrupted material flow from raw materials to crude product.
05 Transcending Ultramicrofactory: The Engineering Form of an Intelligent Unit
This process is implemented on a skid-mounted transcending microfactory developed by Ossen, featuring a modular two-unit architecture. The five-step reaction sequence and inline workup are encapsulated within two combinable skid packages.
Segmented modular integration delivers flexibility and reproducibility. When the target molecule changes or reaction steps are added or removed, the entire line does not need to be rebuilt. Instead, modules can be replaced and reconfigured section by section. Multi-step cascade operation with no intermediate separation brings a marked improvement in space-time efficiency compared with batch routes. Continuous operation of the high-risk azidation reaction transforms a step that was previously difficult to scale due to safety limits into a continuous process with milliliter-scale inventory.
The theme of this annual conference is “Continuous Intelligent Manufacturing for New Chemical Industry”. Featuring a domestically built skid-mounted unit in the academician’s plenary talk signals that continuous processing for Chinese pharmaceutical manufacturing has moved beyond feasibility demonstration. It has become an engineering capability that can be constructed, delivered and reused. Continuous intelligent manufacturing reshapes the R&D paradigm at the reaction level and provides AI4S with a stable source of high-quality experimental data. With this deployable continuous equipment platform, closed-loop iteration between AI and chemical experiments can break the inherent limits of batch workflows, enabling large-scale implementation of intelligent R&D.
