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

Benchmarking protein classification algorithms via supervised cross-validation.

Attila Kertész-Farkas1, Somdutta Dhir, Paolo Sonego

  • 1Research Group on Artificial Intelligence of the Hungarian Academy of Sciences and University of Szeged, Aradi vértanúk tere 1., H-6720 Szeged, Hungary. kfa@inf.u-szeged.hu

Journal of Biochemical and Biophysical Methods
|July 3, 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

SeedMatExplorer: the transcriptome atlas of Arabidopsis seed maturation.

BMC plant biology·2026
Same author

Label-Free Quantification in the Crux Toolkit.

Journal of proteome research·2026
Same author

Genomic and functional analysis of stress-responsive prophages in <i>Lactobacillus helveticus</i>.

Frontiers in microbiology·2026
Same author

Sequencing DNA methylation and hydroxymethylation at co-occurring chromatin features.

Nature communications·2026
Same author

Translational landscape during seed germination revealed by ribosome profiling.

The Plant journal : for cell and molecular biology·2026
Same author

Commentary on 'Untargeted CUT&Tag reads are enriched at accessible chromatin and restrict identification of potential G4-forming sequences in G4-targeted CUT&Tag experiments'.

Nucleic acids research·2025
Same journal

Proceedings of the 12th International Symposium Of Chemistry, Miercurea Ciuc (Csíkszereda), Romania, October 5-8, 2006; and proceedings of the 7th International Symposium and Summer School on Bioanalysis, Pécs, Hungary, June 10-17, 2007.

Journal of biochemical and biophysical methods·2008
Same journal

12th International Symposium Of Chemistry, Miercurea Ciuc (Csíkszereda), Romania, 5th to 8th of October, 2006. 7th International Symposium and Summer School on Bioanalysis, Pécs, Hungary, 10th to 17th of June, 2007.

Journal of biochemical and biophysical methods·2008
Same journal

Optimized HPLC method for tramadol and O-desmethyl tramadol determination in human plasma.

Journal of biochemical and biophysical methods·2008
Same journal

The alternating current polarographic behavior and determination of lansoprazole and omeprazole in dosage forms and biological fluids.

Journal of biochemical and biophysical methods·2008
Same journal

Continuous-flow complete-mixing system for assessing the effects of environmental factors on colony-level coral metabolism.

Journal of biochemical and biophysical methods·2008
Same journal

Utilization of fluorescence tracer in hyperinsulinemic-euglycemic clamp test in mice.

Journal of biochemical and biophysical methods·2008
See all related articles

Developing reliable protein classification algorithms requires realistic benchmark datasets. This study introduces supervised cross-validation strategies and new benchmark datasets, offering more accurate performance estimates for machine learning models in protein science.

Area of Science:

  • Computational biology
  • Bioinformatics
  • Machine learning in genomics

Background:

  • Protein classification algorithm development is challenged by the vast and uneven distribution of protein families.
  • Traditional cross-validation methods may overestimate algorithm generalization to novel protein subtypes.
  • Existing benchmark datasets may not adequately represent the complexity of the protein universe.

Purpose of the Study:

  • To extend supervised cross-validation principles to diverse protein databases.
  • To design standardized, hierarchical benchmark datasets for protein classification.
  • To provide more realistic performance estimates for protein classification algorithms.

Main Methods:

  • Utilized hierarchical classification trees for supervised cross-validation strategies.

Related Experiment Videos

  • Developed benchmark datasets at various conceptual hierarchy levels using graph-theoretic distance.
  • Constructed reduced-size model datasets using a combination of supervised and random sampling.
  • Evaluated machine learning algorithms (e.g., SVM, Random Forests) and comparison algorithms (e.g., BLAST, DALI).

    • Main Results:

    • Added over 3000 new classification tasks to the protein classification benchmark collection.
    • The new datasets include protein sequence, structure, and DNA sequence data.
    • The developed benchmark datasets yield more realistic performance estimates compared to random cross-validation.

    Conclusions:

    • Supervised cross-validation offers a robust framework for creating reliable protein classification benchmarks.
    • The new benchmark datasets facilitate more accurate evaluation of protein classification algorithms.
    • This work enhances the development and testing of machine learning models for protein science.