Ranking, selecting, and prioritising genes with desirability functions.

Area of Science:

• Genomics
• Bioinformatics
• Biostatistics

Background:

• Functional genomics experiments generate large datasets of differentially expressed genes.
• Traditional gene selection methods rely on arbitrary thresholds (p-values, fold-changes) and ad hoc criteria, which can be inefficient and overlook relevant genes.
• Current methods do not adequately account for measurement uncertainty or integrate diverse data sources.

Purpose of the Study:

• To introduce and demonstrate the utility of desirability functions for a more robust and integrated approach to gene selection in functional genomics.
• To provide a flexible framework for ranking and prioritizing genes based on multiple criteria, including experimental results and database information.
• To address the limitations of conventional threshold-based gene selection methods.

Main Methods:

• Desirability functions are employed to map various selection criteria (p-values, fold-changes, etc.) onto a continuous 0-1 scale, where 1 represents maximum desirability.
• Multiple criteria are combined to calculate an overall desirability score for each gene.
• A breast cancer microarray dataset is used to illustrate the application of the desirability function approach for gene prioritization.

Main Results:

• The desirability function approach allows for the integration of multiple data types and experimental outcomes into a single ranking metric.
• This method provides a more nuanced prioritization of genes compared to traditional binary thresholding.
• The approach demonstrated effective gene selection and prioritization in a real-world breast cancer dataset.

Conclusions:

• Desirability functions offer an efficient and flexible alternative to traditional gene selection methods in functional genomics.
• This approach enhances the ability to identify and prioritize relevant genes by comprehensively integrating diverse data and accounting for uncertainty.
• The desiR R package is available to facilitate the implementation of this method for researchers.