Related Experiment Video
Updated: Jul 19, 2025
08:51
Author Spotlight: Integrated Multi-Omics Analysis for Unveiling Multicellular Immune Signatures in Clinical Heart Attack Cohorts
Published on: September 20, 2024
1.3K
MapperPlus: Agnostic clustering of high-dimension data for precision medicine
Esha Datta1, Aditya Ballal2, Javier E López3
1Department of Mathematics, Graduate Group in Applied Mathematics, University of California, Davis, United States of America.
PLOS Digital Health
|August 9, 2023
Summary
Precision medicine aims to tailor treatments by classifying patients into subgroups. MapperPlus, a novel pipeline, effectively identifies these patient clusters, outperforming traditional methods and revealing distinct survival rates in a pediatric stem cell transplant dataset.
Area of Science:
- Computational biology
- Data science
- Precision medicine
Background:
- Precision medicine requires patient stratification into subgroups for tailored treatments.
- Identifying distinct patient clusters is crucial but challenging due to unknown cluster numbers, validity assessment needs, and clinical interpretability.
- Existing clustering methods often struggle with these complexities.
Purpose of the Study:
- To introduce MapperPlus, a novel unsupervised clustering pipeline designed to address key challenges in patient stratification.
- To automatically detect disjoint patient subgroups using an extended topological approach combined with random-walk algorithms.
- To demonstrate the superior performance and predictive power of MapperPlus compared to traditional clustering methods.
Main Methods:
- Developed MapperPlus, an unsupervised clustering pipeline extending the topological Mapper technique.
- Integrated two random-walk algorithms to enhance subgroup detection.
- Validated MapperPlus on diverse public datasets, including medical and non-medical data.
- Applied MapperPlus to a pediatric stem cell transplant dataset to assess its predictive power in stratifying patients by survival rates.
Main Results:
- MapperPlus effectively identifies disjoint patient subgroups.
- The pipeline outperforms traditional agnostic clustering methods in key accuracy and performance metrics.
- MapperPlus demonstrated significant predictive power by stratifying pediatric stem cell transplant patients into clusters with distinct survival rates.
- The number of clusters was automatically detected, addressing a key challenge in patient stratification.
Conclusions:
- MapperPlus is a novel and effective unsupervised clustering pipeline for patient stratification in precision medicine.
- The software successfully addresses critical challenges including unknown cluster numbers, cluster validity assessment, and clinical interpretability.
- MapperPlus offers a significant advancement over traditional methods, providing valuable insights into patient subgroups and their clinical outcomes.
- The open-source availability of MapperPlus facilitates its adoption and further development in medical research.
