Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Videos

Predicting patient survival from microarray data by accelerated failure time modeling using partial least squares and

Susmita Datta1, Jennifer Le-Rademacher, Somnath Datta

  • 1Department of Bioinformatics and Biostatistics, University of Louisville, Louisville, Kentucky 40202, USA.

Biometrics
|April 24, 2007
PubMed
Summary
This summary is machine-generated.

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Identifying Single-Cell Expression Quantitative Trait Loci Using a Bootstrap Penalized Hurdle Model.

Genes·2026
Same author

A Novel Bioinformatics Pipeline and a Machine-Learning Approach for Antimicrobial Resistance Phenotypic Prediction.

Bioinformatics and biology insights·2026
Same author

Unpacking X (formerly Twitter) discourse on fluoride and related topics during the 2024 US presidential election.

Journal of the American Dental Association (1939)·2026
Same author

Pembrolizumab with or without chemotherapy among older adults with advanced lung adenocarcinoma: a national, nonrandomized open-label phase II trial (Alliance A171901).

Journal of the National Cancer Institute·2026
Same author

A lifecycle governance and learning health system framework for trustworthy, generalizable, and sustainable human-ai partnership in clinical practice: Lessons from the asthma-guidance and prediction system (A-GPS).

Journal of the National Medical Association·2026
Same author

Deciphering sepsis molecular subtypes using large-scale data to identify subtype-specific drug repurposing.

bioRxiv : the preprint server for biology·2026

Predicting cancer patient survival from gene expression data is improved using penalized regression methods like LASSO and PLS, especially with censored data. Mean imputation best handled censored data, identifying key cancer-related genes.

Area of Science:

  • Bioinformatics
  • Genomics
  • Statistical modeling

Background:

  • Predicting cancer patient survival from gene expression data is crucial for treatment planning.
  • High-dimensional microarray data presents challenges due to a large number of covariates and few samples.
  • Censored survival data, common in clinical studies, requires specialized statistical handling.

Purpose of the Study:

  • To evaluate the performance of Partial Least Squares (PLS) and Least Absolute Shrinkage and Selection Operator (LASSO) for predicting cancer patient survival times using gene expression data.
  • To investigate the impact of different censoring handling methods (reweighting, mean imputation, multiple imputation) on prediction accuracy.
  • To compare PLS and LASSO performance on high-dimensional microarray data with limited samples.

Main Methods:

Related Experiment Videos

  • Linear regression modeling of log-transformed survival times.
  • Application of Partial Least Squares (PLS) and Least Absolute Shrinkage and Selection Operator (LASSO) with modifications for censored data.
  • Simulation studies to assess performance of censoring imputation methods (reweighting, mean, multiple imputation) and compare with uncensored data.

Main Results:

  • LASSO generally outperformed PLS in prediction error, particularly when many irrelevant variables were present in the gene expression data.
  • For a moderate sample size (100 samples, 10,000 covariates), LASSO provided better predictions than a model with no covariates.
  • Mean imputation demonstrated the best performance, closely matching the results obtained from analyzing complete, uncensored data.

Conclusions:

  • LASSO and PLS are effective methods for predicting cancer survival from gene expression data, even with high dimensionality and censoring.
  • Mean imputation is a robust strategy for handling right-censored survival data in gene expression analysis.
  • Reanalysis of lung cancer data using mean-imputed PLS and LASSO successfully identified known cancer-related genes.