PathMED: an R toolkit for single-sample molecular scoring and machine learning with omics data

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Molecular scoring aids in analyzing pathway-level functional alterations in omics data.
• Molecular scores offer biological interpretability and generalizability across datasets for machine learning.
• A lack of unified tools hinders the use of diverse molecular scoring methods for model training and prediction.

Purpose of the Study:

• To develop pathMED, an R/Bioconductor package that unifies various molecular scoring methods.
• To provide a machine learning module within pathMED for training and testing predictive models.
• To demonstrate pathMED's utility in analyzing omics data for clinical outcome prediction and pathway association.

Main Methods:

• Development of the pathMED R/Bioconductor package.
• Integration of diverse molecular scoring methods into a unified framework.
• Implementation of a machine learning module for model training and prediction using molecular scores.
• Application of gene set dissection for pathway-level analysis.

Main Results:

• pathMED unifies multiple molecular scoring methods and includes a machine learning module.
• Demonstrated generalizability of machine learning models across transcriptomic and proteomic scores.
• Successfully predicted breast cancer treatment response and identified associated pathways.
• Showcased the benefit of gene set dissection for resolving pathway-level disease heterogeneity.

Conclusions:

• pathMED provides a unified and user-friendly framework for molecular scoring and machine learning applications in omics data analysis.
• The package facilitates the prediction of clinical outcomes and the identification of key biological pathways.
• pathMED supports advanced analyses like gene set dissection to uncover complex biological insights.