Quantifying tumor specificity using Bayesian probabilistic modeling for drug and immunotherapeutic target discovery

Area of Science:

• Computational biology
• Genomics
• Drug discovery

Background:

• Therapeutic strategy development for diseases like cancer involves costly and potentially dangerous testing for off-target effects.
• Identifying disease-specific targets is crucial for developing safer and more effective treatments.

Purpose of the Study:

• To develop a systematic method for learning gene and protein disease specificity using molecular measurements from healthy tissues.
• To improve the prioritization of therapeutic targets in oncology drug development.

Main Methods:

• A hierarchical Bayesian modeling approach (BayesTS) was used to integrate protein and gene expression data.
• The method incorporates tunable parameters to adjust for tissue essentiality.
• BayesTS was extended to include splicing antigens and combinatorial target pairs.

Main Results:

• BayesTS outperforms existing strategies in defining therapeutic targets.
• The method successfully nominated previously unrecognized therapeutic targets.
• Extensions of BayesTS yielded more specific targets for therapeutic applications.

Conclusions:

• BayesTS offers a robust computational framework for identifying safer and more effective therapeutic targets.
• This approach can facilitate improved target prioritization in oncology, potentially leading to better drug development outcomes.
• The method's flexibility allows for the integration of diverse molecular data modalities.