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

Independent component analysis yields chemically interpretable latent variables in multivariate regression.

Mats G Gustafsson1

  • 1Uppsala University, Department of Engineering Sciences, Box 528, 751 20 Uppsala, Sweden. Mats.Gustafsson@signal.uu.se

Journal of Chemical Information and Modeling
|September 27, 2005
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

Exhaustive in vitro evaluation of the 9-drug cocktail CUSP9 for treatment of glioblastoma.

Computers in biology and medicine·2024
Same author

A Novel Multiplex Based Platform for Osteoarthritis Drug Candidate Evaluation.

Annals of biomedical engineering·2020
Same author

COMBSecretomics: A pragmatic methodological framework for higher-order drug combination analysis using secretomics.

PloS one·2020
Same author

An ex vivo tissue model of cartilage degradation suggests that cartilage state can be determined from secreted key protein patterns.

PloS one·2019
Same author

COMBImage2: a parallel computational framework for higher-order drug combination analysis that includes automated plate design, matched filter based object counting and temporal data mining.

BMC bioinformatics·2019
Same author

COMBImage: a modular parallel processing framework for pairwise drug combination analysis that quantifies temporal changes in label-free video microscopy movies.

BMC bioinformatics·2018

Independent component analysis (ICA) enables chemically interpretable latent variables in multivariate regression. This method offers accurate predictions and enhanced interpretability over traditional techniques like principal component regression.

Area of Science:

  • Chemometrics
  • Data Analysis
  • Multivariate Statistics

Background:

  • Multivariate regression is crucial for analyzing complex datasets.
  • Latent variables (LVs) in regression often lack clear chemical meaning.
  • Interpretability of LVs is vital for scientific understanding and application.

Purpose of the Study:

  • To investigate the application of independent component analysis (ICA) for extracting chemically interpretable latent variables (LVs) in multivariate regression.
  • To introduce and evaluate novel ICA-based algorithms against established regression methods.
  • To assess the practical limitations and performance of ICA in regression under various conditions.

Main Methods:

  • Development and implementation of two novel algorithms based on independent component analysis (ICA).

Related Experiment Videos

  • Comparison of ICA-based methods with classical techniques: principal component regression (PCR) and partial least-squares regression (PLSR).
  • Utilized simulated data to evaluate prediction accuracy and LV interpretability.

    • Main Results:

    • All compared methods, including ICA, PCR, and PLSR, achieved accurate predictions on simulated data.
    • Only the independent component analysis (ICA)-based regression methods successfully yielded statistically independent and chemically interpretable latent variables (LVs).
    • Simulations highlighted practical considerations for ICA, including sensitivity to underlying assumptions, sample size, and measurement noise.

    Conclusions:

    • Independent component analysis (ICA) provides a powerful approach for obtaining chemically meaningful latent variables in multivariate regression.
    • ICA-based regression offers superior interpretability compared to traditional methods like PCR and PLSR, while maintaining predictive accuracy.
    • Understanding the limitations of ICA is essential for its effective application in real-world chemical data analysis.