Development and Application of Computational Methods for Rational Design of Molecular Glue

Molecular glue is a positive regulator that enhances protein-protein interactions and has been discovered in natural plant hormones, antibiotics, and anti-tumor drugs. Thalidomide and its analogs represent typical examples of molecular glue used as drugs. In cells, these compounds can bind to E3 ubiquitin ligase CRBN and thereby induce the degradation of target proteins. This discovery has sparked significant interest in using molecular glue to induce protein degradation and as a potential therapeutic approach. However, many discoveries of molecular glue still depend on serendipity or experimental screening. There are two major challenges when it comes to discovering molecular glue targets and predicting their pharmacological effects. Firstly, the lack of effective methods for identifying potential targets makes it difficult to verify whether a particular protein can be targeted by a molecular glue, and subsequently design a molecular glue for that target. Secondly, even when a molecular glue compound structure is confirmed, it remains difficult to model the ternary complex induced by the molecular glue and evaluate its degradation efficiency. Overall, these challenges indicate the need for more sophisticated approaches to enable the discovery and application of molecular glues in drug development.

Fig. 1 The identification of potential substrates for the CRBN. A) The PPI-Miner pipeline was designed to identify potential substrates of CRBN by incorporating structural motifs. B) Ternary complex structures of three reported CRBN/molecular glue/substrates modeled by PPI-Miner.

In analyzing the complex structures of CRBN and its substrates, we determined that the interaction between CRBN and its substrates is mediated by a structural motif. To this end, we developed a computational method called PPI-Miner that enables rapid and accurate identification of protein-protein interactions based on these structural motifs. By utilizing this method, we identified several potential substrates of CRBN and created a high-quality target database. Among them, multiple proteins confirmed as molecular glue targets in recent years (as shown in Fig. 1). In addition, PPI-Miner can be used for discovering other motif-mediated protein-protein interactions, as well as searching for protein scaffolds for designing mini-protein binders.