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 protein secondary structure using neural net and statistical methods.

P Stolorz1, A Lapedes, Y Xia

  • 1Theoretical Division, Los Alamos National Laboratory, NM 87545.

Journal of Molecular Biology
|May 20, 1992
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

Automatic segmentation of the glenohumeral cartilages from magnetic resonance images.

Medical physics·2016
Same author

[Analysis of microRNA regulatory network in cochlear hair cells with oxidative stress injury].

Zhonghua er bi yan hou tou jing wai ke za zhi = Chinese journal of otorhinolaryngology head and neck surgery·2016
Same author

Calpain-activated mTORC2/Akt pathway mediates airway smooth muscle remodelling in asthma.

Clinical and experimental allergy : journal of the British Society for Allergy and Clinical Immunology·2016
Same author

Loss of PRDM1/BLIMP-1 function contributes to poor prognosis of activated B-cell-like diffuse large B-cell lymphoma.

Leukemia·2016
Same author

Transcription factor 7-like 1 dysregulates keratinocyte differentiation through upregulating lipocalin 2.

Cell death discovery·2016
Same author

Low natural killer (NK) cell counts in peripheral blood adversely affect clinical outcome of patients with follicular lymphoma.

Blood cancer journal·2016

Neural networks and Bayesian methods show similar protein secondary structure prediction accuracy. Despite sophisticated algorithms, significant progress remains limited, highlighting the need for better biophysical understanding.

Area of Science:

  • Computational Biology
  • Bioinformatics
  • Machine Learning

Background:

  • Predicting protein secondary structure from primary sequence is a fundamental challenge in bioinformatics.
  • Existing methods, including Bayesian and neural network approaches, have shown varying degrees of success.

Purpose of the Study:

  • To compare the efficacy of neural network methods and Bayesian statistical methods for protein secondary structure prediction.
  • To explore novel objective functions for training neural networks in this context.

Main Methods:

  • Comparison of Bayesian statistical models with neural network architectures.
  • Derivation of a neural network formalism where output neurons represent conditional probabilities.
  • Introduction and application of mutual information as an objective function for training.

Related Experiment Videos

Main Results:

  • Bayesian methods, despite a simplifying independence assumption, achieve accuracy comparable to more complex methods.
  • Neural networks trained with mutual information show improved prediction of helix and sheet structures compared to mean-square error.
  • A combined objective function yields a marginal accuracy improvement to 64.4%.

Conclusions:

  • Current sophisticated algorithms offer limited improvement over simpler models for protein secondary structure prediction.
  • Further advancements necessitate a deeper understanding of protein biophysics rather than solely algorithmic improvements.