From undruggable to degradable: A deep learning-enabled framework for precision orthopaedic protein degradation

Area of Science:

• Orthopaedic research
• Computational biology
• Drug discovery

Background:

• Musculoskeletal disorders like osteoarthritis and bone metastases involve "undruggable" targets within bone niches.
• Targeted Protein Degradation (TPD) is a promising therapeutic strategy but faces challenges in orthopaedic applications.
• Barriers include poor degrader physicochemical properties and lack of bone-specific E3 ligases.

Purpose of the Study:

• To present a framework integrating deep learning (DL) with multimodal omics for orthopaedic TPD.
• To explore how DL architectures can redefine the orthopaedic TPD pipeline.
• To overcome limitations in translating TPD to musculoskeletal disorder treatment.

Main Methods:

• Utilizing deep learning architectures such as geometric deep learning, protein language models, and generative design.
• Employing computational strategies for target prioritization and E3 ligase identification.
• Applying a "penetrability-first" design approach for optimising degrader properties.

Main Results:

• Identification of computational methods to prioritize cryptic pockets on skeletal targets.
• Discovery of tissue-restricted E3 ligases using single-cell transcriptomics to minimize toxicity.
• Optimization of degrader permeability for navigating dense bone and cartilage matrices.

Conclusions:

• A computational roadmap bridges the gap between prediction and skeletal pathobiology.
• This approach facilitates a shift from palliative care to mechanism-based microenvironmental reprogramming.
• The framework accelerates the development of precision degraders for joint degeneration, osteoporosis, and bone tumors.