Probabilistic prioritization of candidate pathway association with pathway score

Area of Science:

• Bioinformatics
• Computational Biology
• Systems Biology

Background:

• Current gene-set enrichment analyses often focus on single sets, lacking a quantitative measure for prioritizing significant pathways.
• Existing methods can be influenced by the number of differentially expressed genes and inter-gene correlations within a set.
• P-values alone are insufficient for ranking pathway significance due to their nature as measures of statistical significance, not strength of association.

Purpose of the Study:

• To develop a robust two-stage procedure for prioritizing gene sets and pathways based on their strength of association.
• To introduce a novel pathway-level score that incorporates all genes within a set and accounts for gene correlations.
• To enable accurate ranking of pathways using a Bayesian logistic regression model.

Main Methods:

• A pathway-level score was developed, incorporating all genes in a set and synchronizing gene correlation directions.
• Rank transformation was applied to the pathway score to enhance inter-sample variation and mitigate gene heterogeneity.
• A Bayesian logistic regression model was employed in the second stage to simultaneously evaluate and rank pathway association strengths based on posterior probabilities.

Main Results:

• The proposed summary pathway score effectively evaluates gene expression within gene sets on a per-sample basis.
• The method demonstrated advantages in including all genes and synchronizing correlation directions, outperforming existing approaches.
• The Bayesian model accurately identified top-ranking pathways, providing a probabilistic evaluation of pathway association.

Conclusions:

• The novel pathway scoring and Bayesian ranking method offers a reliable approach for prioritizing biological pathways.
• This method provides a quantitative measure for pathway association strength, aiding in the identification of key biological pathways.
• The ranked pathway lists can guide future research, resource allocation, and the identification of potential therapeutic targets.