Protocol for assessing distances in pathway space for classifier feature sets from machine learning methods

Aug 5, 2026

Upload by Miles

Protocol describing how to reconcile differences in machine-learning (ML) classifier features derived from gene-expression data. The method loads the PathwaySpace R package, prepares analysis inputs, and builds pathway-space visualizations using density plots of gene sets. It further provides procedures to test whether seemingly distinct feature sets are related within pathway space, using a pathway-distance metric to quantify statistical distances between gene-list relationships for subtype-relevant biology.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages12

Reading Time30 min

AccessFree to download

Overview

Protocol goals and rationale

Pathway space construction

Loading PathwaySpace and preparing inputs

Creating density plots of gene sets

Pathway distance assessment

Building and applying the pathway distance metric

Testing relationships between gene sets in pathway space

UC Santa Cruz UC Santa Cruz Previously Published Works Title Protocol for assessing distances in pathway space for classifier feature sets from machine learning methods. Permalink [https://escholarship.org/uc/item/7r6921kj](https://escholarship.org/uc/item/7r6921kj) Journal STAR Protocols, 6(2) Authors Tercan, Bahar Apolonio, Victor Chagas, Vinicius et al. Publication Date 2025-03-18 DOI 10.1016/j.xpro.2025.103681 Peer reviewed [eScholarship.org](eScholarship.org) Powered by the California Digital Library University of California ll OPEN ACCESS Protocol Protocol for assessing distances in pathway space for classiﬁer feature sets from machine learning methods Bahar Tercan, Victor H. Apolonio, Vinicius S. Chagas, ..., Peter W. Laird, Andrew D. Cherniack, Mauro A.A. Castro btercan@systemsbiology. org (B.T.) [peter.laird@vai.org](peter.laird@vai.org)[ ](peter.laird@vai.org)(P.W. L.) achernia@broadinstitute. org (A. D.C.) [mauro.castro@ufpr.br](mauro.castro@ufpr.br)[ ](mauro.castro@ufpr.br)(M.A.A.C.) Highlights Protocoltotestwhether distinct gene sets reﬂect related biology Steps for building a pathway space for graph-or networkbased distance analysis Instructions for calculating a pathway distance metric for any pair of gene sets Guidance on exploring relationships between gene sets in pathway space Tercan et al., STAR Protocols 6, 103681 June 20, 2025 ª 2025 The Authors. Published by Elsevier Inc. [https://doi.org/10.1016/](https://doi.org/10.1016/)[ ](https://doi.org/10.1016/)[j.xpro.2025.103681](j.xpro.2025.103681) ll OPEN ACCESS Protocol Protocol for assessing distances in pathway space forclassiﬁer feature sets from machine learning methods Bahar Tercan,1, 15, 16,* Victor H. Apolonio,2, 15 Vinicius S. Chagas,2 Christopher K. Wong,3 JordanA. Lee,4 Christina Yau,5,6 Christopher C. Benz,6 Joshua M. Stuart,3 Brian J. Karlberg,4 Kyle Ellrott,4 Jasleen K. Grewal,7, 13 Steven J. M. Jones,7 The Cancer Genome Atlas Analysis Network, Jean C. Zenklusen,8 A. Gordon Robertson,7, 14 Peter W. Laird,9,* Andrew D. Cherniack, 10, 11, 12,* and Mauro A.A. Castro2, 17,* 1Institute of Systems Biology, 401 Terry Avenue North, Seattle, WA 98109, USA 2Bioinformatics and Systems Biology Laboratory, Federal University of Parana´, Curitiba, PR 81520-260, Brazil 3UC Santa Cruz Genomics Institute and Department of Biomolecular Engineering, Santa Cruz, CA 95060, USA 4Oregon Health and Science University, Portland, OR 97239, USA 5University of California, San Francisco, Department of Surgery, San Francisco, CA 94158, USA 6Buck Institute for Research on Aging, Novato, CA 94945, USA 7Canada’s Michael Smith Genome Sciences Centre, BC Cancer, Vancouver, BC, Canada 8Center for Cancer Genomics, National Cancer Institute, Bethesda, MD 20892, USA 9Department of Epigenetics, Van Andel Institute, Grand Rapids, MI 49503, USA 10The Broad Institute of Harvard and MIT, Cambridge, MA 02142, USA 11Department of Medical Oncology, Dana-Farber Cancer Institute, Boston, MA 02215, USA 12Harvard Medical School, Boston, MA 02115, USA 13Present address: NVIDIA Corporation, Santa Clara, CA, USA 14Present address: Dxige Research Inc., Courtenay, BC V9N 1C2, Canada 15These authors contributed equally 16Technical contact 17Lead contact *Correspondence: [btercan@systemsbiology.org](btercan@systemsbiology.org) (B.T.), [peter.laird@vai.org](peter.laird@vai.org) (P.W. L.), [achernia@broadinstitute.org](achernia@broadinstitute.org) (A. D.C.), [mauro.castro@ufpr.br](mauro.castro@ufpr.br) (M.A.A.C.) [https://doi.org/10.1016/j.xpro.2025.103681](https://doi.org/10.1016/j.xpro.2025.103681) SUMMARY As genes tend to be co-regulated as gene modules, feature selection in machine learning (ML) on gene expression data can be challenged by the complexity of gene regulation. Here, we present a protocol for reconciling differences in classiﬁer features identiﬁed using different ML approa

FAQs

What problem does the PathwaySpace protocol address?

It addresses how to compare pathway distances between gene sets and determine whether distinct ML-derived classifier feature sets reflect related biology.

What are the main steps in executing the protocol?

Load the PathwaySpace R package, prepare inputs for pathway-space analysis, construct pathway-space visualizations (including density plots), and then run pathway-distance procedures for gene-set comparisons.

How does the protocol evaluate relationships between gene sets?

It uses a pathway-distance metric and provides procedures to test whether apparently distinct feature sets are related in pathway space, supporting interpretation for pathway- and subtype-relevant biology.

Aug 5, 2026

Upload by Miles