Antibodies are a rapidly expanding class of therapeutic agents due to their alluring drug-like characteristics, such as high target selectivity and negligible immunogenic effects. The initial lead molecule is often discovered when developing an antibody medication. Current methods for its development often involve screening huge libraries of random antibody variants against a target antigen to haphazardly search through enormous combinatorial sequence spaces. Initial discovery is often accomplished using methods like yeast display, phage display, vaccination combined with hybridoma screening, or B-cell sequencing, which is then followed by subsequent molecular development. These techniques take a lot of time and resources, lack control over the characteristics of the resultant antibodies, and frequently result in poor leads. Instead of screening and creating lead compounds, generative artificial intelligence (AI) might be used to generate de novo antibodies in a zero-shot and controllable manner, which would significantly cut down on the time and resources needed for therapeutic antibody production.
Artificial Inteligence
Zero-shot antibody design
Given the availability of extensive protein sequence and structural databases that can be utilized for model training, the use of AI approaches to antibody development and, more broadly, protein therapeutic design, appears intriguing. Recent research has demonstrated that models built on these data could be applied to the de novo construction of specific protein classes.
A study was carried out by a group of scientists to demonstrate zero-shot antibody design via an experiment featuring a wet lab technique. This entails developing an antibody to link to an antigen without any evidence of existing binders to the antigen beforehand. Using engineering concepts to create whole new antibodies is another way to put it. The previous studies screened hundreds to thousands of protein designs, which is a two- to four-order-of-magnitude smaller number of proteins than the ones validated in this study. Also, the design of the complementary determining regions (CDRs) is crucial for determining how well an antibody works and interacts with the antigen. It was demonstrated that heavy chain CDRs (HCDR) can be created using generative AI techniques employing trastuzumab and its target antigen, HER2, as a model system as a first step towards an entirely de novo antibody design. All antigens that bind to HER2 or its homologs were taken out of the training set. Using a patented Activity Specific Cell-Enrichment (ACE) test, 440,000 different HCDR3 variations of trastuzumab were created de novo, and binding to HER2 was checked. Among the various designs, 421 binders were functionally confirmed using surface plasmon resonance (SPR), and we calculated the presence of about 4,000 binders in total. The created binders are not only very diverse and distinct from anything previously seen in structural antibody databases or enormous datasets of known antibodies, but they also possess sequence uniqueness in comparison to those identified in the training dataset. Additionally, the approach’s extensibility was demonstrated by creating and evaluating binding molecules for two additional antigens: vascular endothelial growth factor A (VEGF-A) and the SARS-CoV-2 spike RBD.
Read Also: Solid Tumor Destruction: A Possible Breakthrough with Car T Cell Immunotherapy
Clinical significance
The work is a significant development for in silico antibody design and has the potential to transform patient access to efficient medicines. The capacity to regulate AI-designed antibodies will allow the generation of unique molecules for particular disease targets, resulting in safer and more effective medicines than would be achievable using conventional development methods.
Conclusion
The speed, quality, and controllability of antibody design could all be considerably improved with the help of generative artificial intelligence (AI). Traditional de novo antibody development requires extensive screening of huge immunological or synthetic libraries, which takes time and resources. However, employing generative AI and high-throughput experimentation, de novo antibody design opens the possibility of rapid drug development for new therapeutic targets.
References
Unlocking de novo antibody design with generative artificial intelligence
FEEDBACK: