Tianwu Tech and Jin赛 Pharma Achieve Global First: AI-Designed Protein Successfully Scaled to 5,000-Liter Industrial Production

June 17 - Matwings Technology ("Matwings Technology") and Changchun GeneScience Pharmaceutical Co., Ltd. ("GenSci") jointly announced that they have made breakthrough progress in designing ultra-alkaline-resistant single-domain antibodies using AI large models, successfully completingThe Development of the World's First Large-Model-Designed Protein Product Entering 5000-Liter Scale-Up Production and Practical Application.

Improving the alkali resistance of proteins has always been a highly challenging task. In the industry, the only truly alkali-resistant affinity ligand protein is Protein A.It took researchers nearly 10 years to improve its alkali resistance to an industrially usable level, but the application scope is still limited to antibodies.

This time, Matwings Technology collaborated with GenSci to fully design through a general large model of protein engineering, combined with a small amount of wet experimental closed-loop iteration verification.In less than a year,The alkali resistance of a common non-alkali-resistant single-domain antibody was increased 4 times and applied in a 5,000-liter scale-up production, marking a significant advancement throughThe maturity of AI large model customized development technology for highly alkali-resistant affinity chromatography media enables the evolution of any single-domain antibody into alkali-resistant affinity chromatography media that can be used industrially. This technology can apply affinity chromatography in the purification of any biomolecule (including GLP-1, cell and gene therapy carrier proteins, AAV viral particles, etc.), replacing multi-step chromatography processes for tag-free proteins.This breakthrough addresses the challenges of high purification difficulty and low production efficiency for this class of molecules, representing a significant technological innovation in downstream processes on a global scale.

Dr. Lei Jin, General Manager of GenSci, stated: "This breakthrough not only marksLow-cost purification using affinity chromatography becomes possible, which also means that artificial intelligence has taken an important step from Science research to Engineering application in the biopharmaceutical field, and has profound significance for improving drug production efficiency, reducing production costs, and promoting the development of new quality productivity."

Growth Hormone (GH) is a peptide hormone secreted by the anterior pituitary gland, playing a crucial role in human growth and development. It promotes human growth and cell proliferation by directly acting on bones, cartilage, and muscles, as well as indirectly stimulating the liver to produce Insulin-like Growth Factor 1 (IGF-1), showing significant effects in treating short stature in children.

As the only long-acting growth hormone currently approved for marketing in China,GenSci JinSaiZeng^®The production method mainly adopts E. coli secretion-type expression technology, utilizing E. coli to efficiently express human growth hormone, which is then manufactured through fermentation, centrifugation, and purification steps.

However, traditional purification methods result in significant growth hormone loss. Therefore, GenSci screened a library of 40 million alpaca single-domain antibodies and successfully identified a single-domain antibody (Single-domain antibody, sdAb) with excellent affinity, also known as a nanobody (Nanobody, Nb), VHH antibody, or camelid antibody—a small, high-affinity protein found in camelid animals. Considering the small size, simple structure, and ease of production of single-domain antibodies, GenSci decided to use a single-domain antibody as the affinity ligand molecule in affinity chromatography to specifically bind growth hormone.

However, "alkali" is the nemesis of the vast majority of proteins, and single-domain antibodies have poor alkali resistance.Most proteins require mild acidic or alkaline conditions to function. Typically, when the pH of the environment is less than 5 or greater than 10, the molecular structure of proteins will be affected. Under more extreme alkaline conditions where the pH exceeds 13, the protein structure will be further disrupted, leading to loss of its function and activity.

During the production process, after using affinity chromatography to purify growth hormone, it is necessary to remove other organic contaminants from the chromatographic media, reduce non-specific adsorption, promote media regeneration, and achieve a disinfecting effect, thereby improving purification efficiency and media reusability. Therefore, a strong alkali (0.5M NaOH, PH 13-14) is required to elute these contaminants. When conventional single-domain antibodies encounter a strongly alkaline environment, the most direct consequence is their drastically shortened service life and increased corporate costs. This is also why affinity chromatography, an excellent purification method, faces challenges in industrial applications.Main Reason for Limited Promotion During Childbirth.

"Matwings Technology Designs Extremely Alkaline-Resistant Single-Domain Antibodies,Compared with other companies or traditional methodsHas Natural Advantages." Dr. Liu Hao, Chief Technology Officer of Matwings Technology, stated: "Our AI protein design general large model, AccelProtein™ (originating from the Pro series general artificial intelligence developed by Professor Hong Liang's team at Shanghai Jiao Tong University), uses a dataset that covers moreExtreme EnvironmentProtein of Lower OrganismsSequence and Structure，and then with the help of strong few-shot learning capabilities, it can quickly capture features related to extreme conditions, which is a capability to predict anti-extreme condition mutants that general large models or traditional methods do not possess."

Traditional methods for improving alkali resistance involve substituting amino acids with weaker alkali tolerance (such as asparagine and glutamine). While this approach is somewhat effective in enhancing the alkali resistance of proteins, its impact is limited. Other types of amino acids and their combinations within proteins can also significantly affect alkali resistance, but there are no universal rules to summarize these effects. As a result, modifications often fall into a situation of relying on random mutations.Making its workload heavy and the process unpredictable, with a difficulty far greater than the modification of indicators such as activity, affinity, and selectivity.

Matwings Technology's AI protein design large model AccelProtein™, based on self-supervised learning of nearly 1 billion protein sequences from various extreme environments in nature, has understood the composition rules of natural proteins and mastered the complex semantic relationships between protein sequences, structures, and their functions.Can directly predict mutants with excellent alkali resistance,This is something that traditional methods and other conventional large models lacking protein datasets in harsh environments cannot achieve.

In the initial design, the AccelProtein™ large model wasSuccessfully designed more than ten without any experimental dataIn terms of alkali resistance, affinity, and thermal stability, etc.Single-point mutants superior to the wild type.After this, AccelProtein™ once again demonstrated its powerful ability to capture feature predictions of multi-point mutations. Many multi-point mutations exhibited stronger alkali resistance, affinity, and thermal stability. Moreover, based on the complex epistatic effects it learned, it even designed sequences where the combination of two negative mutations turned into a positive mutation — something that traditional design methods could never achieve.

Traditional rational design in protein engineering tends to automatically dismiss negative mutations, as human cognition struggles to comprehend how negative sites can improve a multi-site mutant through epistatic effects. In contrast, large models, which are based on protein semantics, can inherently understand these epistatic effects and cleverly incorporate negative sites into mutants to make the "semantics smoother" — thereby enhancing protein function.Large models can ingeniously utilize negative mutations, representing a fundamental innovation across the entire field, significantly enhancing the design capabilities and imaginative scope of protein engineering.。

4MonthsAfter a period of time, the single-domain antibody designed by AccelProtein™ large model was proven in the protein pilot phase: after being treated with 0.5M NaOH for 24 hours, the breakage ratio decreased from 60% (wild type) before modification to 15%, which means enhanced resistance.Alkalinity increased 4 times; The binding ability to growth hormone is twice that of the original; the thermal stability has increased by 8°C compared to before the modification.

After the single-domain antibody was conjugated to the affinity filler and prepared into a chromatography column, the remaining dynamic loading capacity showed that, compared with the wild-type single-domain antibody 14#, the AI-designed protein molecule had significantly improved. For instance, the loading capacity of M74 after 24 hours of alkali treatment was 56.21, which was nearly four times higher than the wild-type’s 15.2.

"The core goal of this cooperation between the two parties is to enhance the alkali resistance of this single-domain antibody, reduce the production cost for enterprises, and improve their production efficiency. The use of single-domain antibodies in affinity chromatography holds great promise and industrial application potential. It can enable customized development of fillers for different substrates, thereby broadening the application scope and enhancing binding specificity," said Dr. Jin Lei.

Dr. Liu Hao believes that as a pioneer in the transition from AI for Science to AI for Engineering, Matwings Technology's collaboration outcomes have demonstrated the industrial application capability of its general AI protein design model AccelProtein™, providing a more cost-effective solution for CMC in biopharmaceuticals.

In the future, Matwings Technology will continue to promote the innovation and development of general artificial intelligence technology in the field of protein engineering to meet the growing needs of the biopharmaceutical industry and contribute more to the cause of human health.