diceR: an R package for class discovery using an ensemble driven approach.

Area of Science:

• Computational biology
• Bioinformatics
• Data science

Background:

• Cluster analysis, an unsupervised learning technique, is vital for identifying patient sub-populations in health and cancer research.
• Traditional methods require upfront selection of algorithms and cluster numbers, posing validation challenges.
• Ensemble clustering enhances generalization and reproducibility of findings.

Purpose of the Study:

• To introduce diceR, an R software package for diverse cluster ensembles.
• To provide tools that guide researchers through cluster analysis while minimizing subjective decision-making.
• To offer a data-agnostic approach applicable beyond biological contexts.

Main Methods:

• Development of the diceR package in R, available on CRAN.
• Implementation of ensemble clustering techniques to address limitations of traditional methods.
• Focus on minimizing subjective choices in the clustering process.

Main Results:

• diceR provides a suite of tools for robust cluster analysis.
• The package facilitates the generalization and reproducibility of patient sub-population discoveries.
• Demonstrates applicability in biological contexts and beyond.

Conclusions:

• diceR aids researchers in objective cluster analysis for health research.
• The package supports improved disease diagnosis, prognosis, and therapy response segmentation.
• Its data-agnostic nature allows for broad application across scientific domains.