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

Systematic learning of gene functional classes from DNA array expression data by using multilayer perceptrons.

Alvaro Mateos1, Joaquín Dopazo, Ronald Jansen

  • 1Bioinformatics Unit, Centro Nacional de Investigaciones Oncologicas (CNIO), 28039, Madrid, Spain.

Genome Research
|November 8, 2002
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

Shaping the extinction burst: Increasing its probability and preventing its emergence across topographies.

Journal of the experimental analysis of behavior·2026
Same author

A decade of progress towards the elimination of hepatitis C: from the National Strategic Plan to Artificial Intelligence.

Gastroenterologia y hepatologia·2026
Same author

Menopause age and hypercholesterolemia comorbidities: a region-wide retrospective cohort study in Andalusia, Spain (2016-2022).

BMJ open·2026
Same author

IMPaCT-Data: A Federated Precision Medicine Infrastructure Associated with Science and Technology in Spain.

Studies in health technology and informatics·2026
Same author

Towards a decentralized future for open-science databases.

Nature genetics·2026
Same author

High-Fidelity Synthetic Data Replicates Clinical Prediction Performance in a Million-Patient Diabetes Cohort.

Advanced science (Weinheim, Baden-Wurttemberg, Germany)·2026
Same journal

A unified analysis of cell type- and trajectory-associated pathways in single-cell data using Phoenix.

Genome research·2026
Same journal

Resf1 is required for proper placental development and configuration of trophoblast cell-specific heterochromatin.

Genome research·2026
Same journal

Telomere-driven replicative crisis is driven by large-scale changes in genomic architecture.

Genome research·2026
Same journal

Spatially informed reference-free cell-type deconvolution for spatial transcriptomics with SpatialCD.

Genome research·2026
Same journal

Spatially resolved profiling of steroid nuclear receptors reveals a role for the disordered N-terminal domains in genome targeting and AP-1 interaction.

Genome research·2026
Same journal

Flexible and scalable inference of spatially varying correlation in spatial transcriptomics with spCorr.

Genome research·2026
See all related articles

Supervised neural networks (SNNs) can now annotate gene function genome-wide. However, biological pathway interconnections, termed the "Borges effect," limit accuracy, necessitating new learning strategies for improved gene function prediction.

Area of Science:

  • Bioinformatics
  • Computational Biology
  • Genomics

Background:

  • Microarray technology enables large-scale gene functional annotation.
  • Supervised learning algorithms, including Support Vector Machines (SVMs), have been used for gene function prediction based on expression signatures.
  • Previous applications were limited to a small number of functional classes.

Purpose of the Study:

  • To comprehensively apply supervised neural networks (SNNs) for genome-wide gene functional annotation across approximately 100 Munich Information Center for Protein Sequences (MIPS) functional classes.
  • To analyze the learnability of these classes and investigate the impact of inter-class biological relationships on machine learning performance.
  • To introduce a novel learning procedure to improve the accuracy of functional annotation.

Main Methods:

Related Experiment Videos

  • Systematic application of supervised neural networks (SNNs) to ~100 MIPS functional classes.
  • Analysis of learnability based on false negative rates.
  • Quantification of the 'Borges effect' using two new numerical indices.
  • Development and application of an iterative learning procedure combining false positives with original classes.

Main Results:

  • Only ~10% of the analyzed MIPS classes were found to be learnable with low false negative rates.
  • The 'Borges effect,' resulting from interconnections among functional classes, confounds unique signature learning.
  • A lower 'Borges effect' in classification systems correlates with better machine learning suitability.
  • The proposed iterative learning procedure converged to a learnable gene set with low false positive/negative rates, biologically related to the target class.

Conclusions:

  • Supervised neural networks offer a powerful approach for genome-wide gene functional annotation.
  • The 'Borges effect' is a significant challenge in machine learning-based functional genomics, highlighting the complexity of biological pathway interactions.
  • The developed iterative learning strategy effectively mitigates the 'Borges effect', improving annotation accuracy and enabling the reconstruction of pathway interactions.
  • This methodology, exemplified by the tricarboxylic acid cycle, provides a robust framework for advancing functional genomics research.