Introduction & Background

Tyrosine kinase 2 (TYK2) is a promising therapeutic target due to its involvement in multiple signaling pathways. In our study, we focused on identifying potential inhibitors for TYK2 using a virtual screening approach. The research incorporated an ultra-large chemical database and benchmarked the performance of our proprietary pipeline against established methods.

Challenge / Objective

The primary objectives were to:

• Identify potential TYK2 inhibitors from the ultra-large ZINC20 database (~900 million compounds).

• Validate the effectiveness of our in-silico screening pipeline by benchmarking it with 20 known active compounds.

• Assess whether our approach could outperform widely used tools, such as AutoDock Vina.

Approach

Our methodology was based on a multi-step process:

• Virtual Screening:

We performed a comprehensive virtual screening of the ZINC20 database, which includes approximately 900 million compounds. This allowed us to capture a wide spectrum of chemical diversity.

• Benchmarking with Known Actives:

To validate our pipeline, we incorporated 20 compounds already known to bind to TYK2 into our initial dataset. These served as benchmarks to evaluate the screening quality.

• Comparison with Established Software:

The performance of our proprietary pipeline was compared to that of AutoDock Vina. Approximately 75% of the known actives were ranked within the top 7% of our final hit list (comprising 500K compounds), while AutoDock Vina achieved only about 43% in the same metric.

• Conformational and Chemical Analysis:

We further analyzed the hits for structural diversity and binding affinity. Notably, some compounds demonstrated even greater binding affinity than Deucravacitinib—the first-in-class selective inhibitor developed by BMS. The analysis revealed novel substructures and scaffolds capable of strong binding within the pseudokinase binding domain pocket.

• Assessment of BBB Permeability:

In addition, some identified compounds were predicted to cross the blood-brain barrier (BBB), which is a critical property for potential treatments of neurodegenerative diseases like Alzheimer’s disease. However, such compounds were relatively scarce in the current shortlist.

Results

• Pipeline Superiority:

Our proprietary screening approach successfully ranked approximately 75% of known actives in the top 7% of a final hit list of 500K compounds, significantly outperforming AutoDock Vina’s 43%. This result validates the superior quality and efficiency of our computational pipeline.

• Diverse and Potent Hits:

The screening uncovered compounds spanning a wide range of chemical classes with remarkable structural diversity and high binding affinity for TYK2. In some cases, the binding affinities exceeded that of Deucravacitinib.

• Novel Chemical Insights:

Conformational and chemical analyses revealed previously unreported substructures and scaffolds that contribute to strong binding within the target’s pseudokinase binding domain.

• BBB Permeability Findings:

Although a subset of compounds was predicted to cross the BBB, these were limited in number. This observation highlights an opportunity to refine the early stages of the screening process.

Impact and Next Steps

• Validation of Computational Approach:

The strong performance metrics not only validate our proprietary pipeline but also underscore its potential for accelerating drug discovery.

• Opportunities for Optimization:

The discovery of structurally diverse and potent compounds, along with novel binding scaffolds, opens new avenues for further optimization of TYK2 inhibitors.

• Future Enhancements:

To increase the identification of compounds with BBB permeability, future experiments will integrate BBB prediction strategies earlier in the screening process and utilize an even larger initial database.