Therapeutic target prediction for orphan diseases integrating genome-wide and transcriptome-wide association studies

Area of Science:

• Computational biology and bioinformatics
• Genomics and transcriptomics
• Drug discovery and development

Background:

• Therapeutic target identification is a critical bottleneck in drug discovery, especially for rare and orphan diseases.
• Existing methods often struggle with diseases lacking extensive genetic or clinical data.
• Understanding disease mechanisms at a molecular level is essential for effective treatment strategies.

Purpose of the Study:

• To introduce TRESOR, a novel disease signature for characterizing functional disease mechanisms.
• To develop machine learning models for predicting inhibitory and activatory therapeutic targets.
• To enhance target identification for rare diseases with limited known targets.

Main Methods:

• Utilized genome-wide association study (GWAS) and transcriptome-wide association study (TWAS) data to define the TRESOR disease signature.
• Developed machine learning algorithms to predict therapeutic targets based on gene perturbation signatures (gene knockdown/overexpression).
• Employed Bayesian optimization with omics-based disease similarities to improve performance for data-poor diseases.

Main Results:

• Successfully identified 4345 inhibitory target candidates across 284 diseases and 4040 activatory target candidates across 151 diseases.
• TRESOR accurately identifies target proteins that counteract disease-specific transcriptome patterns.
• Promising targets were validated using independent cohorts, demonstrating the robustness of the predictions.

Conclusions:

• The TRESOR signature and associated machine learning methods provide a powerful framework for therapeutic target identification.
• This approach significantly aids in discovering targets for rare and orphan diseases.
• The methods can also facilitate a deeper understanding of relationships between different diseases.