|
چکیده
|
The integration of deep learning and bioinformatics has fundamentally transformed biomolecule discovery and targeted ligand design via phage display. Tools such as AlphaFold and next-generation sequencing (NGS) analyses have enabled the rapid identification of protein variants with novel function. Advanced models like LSTM and Transformer architectures optimize binding site affinity and molecular stability, reducing time and cost in drug development. Even in libraries with limited diversity, these models can propose effective modifications to fine-tune binding sites and enhance ligand stability .Despite challenges such as phage display replication bias and data complexity, modern statistical methods like Gaussian processes and more precise selection criteria (based on positive-to-negative selection ratios) have facilitated the construction of focused libraries. Future research directions include improving biopanning methods to better eliminate off-target variants, employing trinucleotide codons for more precise library design, implementing active learning with real-time laboratory feedback, extending these technologies to applications such as IgG antibody and TCR design, integrating multi omics data for a more comprehensive understanding, and ultimately conducting long-term, rigorous evaluations in animal models and clinical studies to conclusively confirm efficacy and safety.
|