A groundbreaking advancement in bioengineering has emerged with the introduction of DeepSCan, a suite of deep learning and artificial intelligence models designed to systematically map the relationships between cell surface display (CSD) sequences and their functional capabilities. This innovative approach addresses the existing gap in understanding how specific CSD sequences correlate with their effectiveness.
Mapping Sequence to Function
DeepSCan enables researchers to reliably identify conserved features of potent CSD elements and iteratively design new, effective CSD modules for mRNA antigen display. The study involved the experimental quantification of surface expression across more than 570 chimeric antigens and the derivation of cell surface translocation strength labels for approximately 310 CSD elements. In total, around 45 independent training datasets were compiled to enhance the model’s accuracy.
Model Training and Validation
Three generations of DeepSCan models were trained and evaluated for performance. Guided by these models, researchers computationally designed 3,700 new CSDs and experimentally validated about 120 of these generative CSDs. Remarkably, they identified 7 generative CSDs that either matched or surpassed the cell surface translocation strength of the most potent naturally occurring CSDs.
Functional Validation Across Cell Types
The enhanced surface displays were validated across multiple cell types and demonstrated functionality in an antigen-specific CAR-T cytotoxicity assay. This validation underscores the potential of the newly designed CSDs in therapeutic applications, particularly in the realm of immunotherapy.
Future Directions
The implications of this research extend beyond mere academic interest, as the findings could pave the way for more effective designs in CSD modules, enhancing their application in various biomedical fields. The systematic approach of DeepSCan offers a robust framework for future explorations in synthetic biology, potentially leading to significant advancements in targeted therapies.
This article was produced by NeonPulse.today using human and AI-assisted editorial processes, based on publicly available information. Content may be edited for clarity and style.
Original source: nature.com
