Neural/Bayes network predictor for inheritable cardiac disease pathogenicity and phenotype

Thomas P Burghardt, Katalin Ajtai

Research output: Contribution to journalArticle

2 Citations (Scopus)

Abstract

The cardiac muscle sarcomere contains multiple proteins contributing to contraction energy transduction and its regulation during a heartbeat. Inheritable heart disease mutants affect most of them but none more frequently than the ventricular myosin motor and cardiac myosin binding protein c (mybpc3). These co-localizing proteins have mybpc3 playing a regulatory role to the energy transducing motor. Residue substitution and functional domain assignment of each mutation in the protein sequence decides, under the direction of a sensible disease model, phenotype and pathogenicity. The unknown model mechanism is decided here using a method combing neural and Bayes networks. Missense single nucleotide polymorphisms (SNPs) are clues for the disease mechanism summarized in an extensive database collecting mutant sequence location and residue substitution as independent variables that imply the dependent disease phenotype and pathogenicity characteristics in 4 dimensional data points (4ddps). The SNP database contains entries with the majority having one or both dependent data entries unfulfilled. A neural network relating causes (mutant residue location and substitution) and effects (phenotype and pathogenicity) is trained, validated, and optimized using fulfilled 4ddps. It then predicts unfulfilled 4ddps providing the implicit disease model. A discrete Bayes network interprets fulfilled and predicted 4ddps with conditional probabilities for phenotype and pathogenicity given mutation location and residue substitution thus relating the neural network implicit model to explicit features of the motor and mybpc3 sequence and structural domains. Neural/Bayes network forecasting automates disease mechanism modeling by leveraging the world wide human missense SNP database that is in place and expanding.

Original languageEnglish (US)
Pages (from-to)19-27
Number of pages9
JournalJournal of Molecular and Cellular Cardiology
Volume119
DOIs
StatePublished - Jun 1 2018

Fingerprint

Virulence
Heart Diseases
Phenotype
Single Nucleotide Polymorphism
Databases
Cardiac Myosins
Ventricular Myosins
Sarcomeres
Mutation
Proteins
Neural Networks (Computer)
Myocardium
Carrier Proteins

Keywords

  • Autonomous motor
  • Cardiac atrial myosin
  • Cardiac myosin binding protein C
  • Cardiac ventricular myosin
  • Dilated cardiomyopathy
  • Hypertrophic cardiomyopathy
  • Inheritable heart disease mechanism
  • Machine learning
  • Restrictive cardiomyopathy

ASJC Scopus subject areas

  • Molecular Biology
  • Cardiology and Cardiovascular Medicine

Cite this

Neural/Bayes network predictor for inheritable cardiac disease pathogenicity and phenotype. / Burghardt, Thomas P; Ajtai, Katalin.

In: Journal of Molecular and Cellular Cardiology, Vol. 119, 01.06.2018, p. 19-27.

Research output: Contribution to journalArticle

@article{41be536f0746497a92c1bcd2eb6afc27,
title = "Neural/Bayes network predictor for inheritable cardiac disease pathogenicity and phenotype",
abstract = "The cardiac muscle sarcomere contains multiple proteins contributing to contraction energy transduction and its regulation during a heartbeat. Inheritable heart disease mutants affect most of them but none more frequently than the ventricular myosin motor and cardiac myosin binding protein c (mybpc3). These co-localizing proteins have mybpc3 playing a regulatory role to the energy transducing motor. Residue substitution and functional domain assignment of each mutation in the protein sequence decides, under the direction of a sensible disease model, phenotype and pathogenicity. The unknown model mechanism is decided here using a method combing neural and Bayes networks. Missense single nucleotide polymorphisms (SNPs) are clues for the disease mechanism summarized in an extensive database collecting mutant sequence location and residue substitution as independent variables that imply the dependent disease phenotype and pathogenicity characteristics in 4 dimensional data points (4ddps). The SNP database contains entries with the majority having one or both dependent data entries unfulfilled. A neural network relating causes (mutant residue location and substitution) and effects (phenotype and pathogenicity) is trained, validated, and optimized using fulfilled 4ddps. It then predicts unfulfilled 4ddps providing the implicit disease model. A discrete Bayes network interprets fulfilled and predicted 4ddps with conditional probabilities for phenotype and pathogenicity given mutation location and residue substitution thus relating the neural network implicit model to explicit features of the motor and mybpc3 sequence and structural domains. Neural/Bayes network forecasting automates disease mechanism modeling by leveraging the world wide human missense SNP database that is in place and expanding.",
keywords = "Autonomous motor, Cardiac atrial myosin, Cardiac myosin binding protein C, Cardiac ventricular myosin, Dilated cardiomyopathy, Hypertrophic cardiomyopathy, Inheritable heart disease mechanism, Machine learning, Restrictive cardiomyopathy",
author = "Burghardt, {Thomas P} and Katalin Ajtai",
year = "2018",
month = "6",
day = "1",
doi = "10.1016/j.yjmcc.2018.04.006",
language = "English (US)",
volume = "119",
pages = "19--27",
journal = "Journal of Molecular and Cellular Cardiology",
issn = "0022-2828",
publisher = "Academic Press Inc.",

}

TY - JOUR

T1 - Neural/Bayes network predictor for inheritable cardiac disease pathogenicity and phenotype

AU - Burghardt, Thomas P

AU - Ajtai, Katalin

PY - 2018/6/1

Y1 - 2018/6/1

N2 - The cardiac muscle sarcomere contains multiple proteins contributing to contraction energy transduction and its regulation during a heartbeat. Inheritable heart disease mutants affect most of them but none more frequently than the ventricular myosin motor and cardiac myosin binding protein c (mybpc3). These co-localizing proteins have mybpc3 playing a regulatory role to the energy transducing motor. Residue substitution and functional domain assignment of each mutation in the protein sequence decides, under the direction of a sensible disease model, phenotype and pathogenicity. The unknown model mechanism is decided here using a method combing neural and Bayes networks. Missense single nucleotide polymorphisms (SNPs) are clues for the disease mechanism summarized in an extensive database collecting mutant sequence location and residue substitution as independent variables that imply the dependent disease phenotype and pathogenicity characteristics in 4 dimensional data points (4ddps). The SNP database contains entries with the majority having one or both dependent data entries unfulfilled. A neural network relating causes (mutant residue location and substitution) and effects (phenotype and pathogenicity) is trained, validated, and optimized using fulfilled 4ddps. It then predicts unfulfilled 4ddps providing the implicit disease model. A discrete Bayes network interprets fulfilled and predicted 4ddps with conditional probabilities for phenotype and pathogenicity given mutation location and residue substitution thus relating the neural network implicit model to explicit features of the motor and mybpc3 sequence and structural domains. Neural/Bayes network forecasting automates disease mechanism modeling by leveraging the world wide human missense SNP database that is in place and expanding.

AB - The cardiac muscle sarcomere contains multiple proteins contributing to contraction energy transduction and its regulation during a heartbeat. Inheritable heart disease mutants affect most of them but none more frequently than the ventricular myosin motor and cardiac myosin binding protein c (mybpc3). These co-localizing proteins have mybpc3 playing a regulatory role to the energy transducing motor. Residue substitution and functional domain assignment of each mutation in the protein sequence decides, under the direction of a sensible disease model, phenotype and pathogenicity. The unknown model mechanism is decided here using a method combing neural and Bayes networks. Missense single nucleotide polymorphisms (SNPs) are clues for the disease mechanism summarized in an extensive database collecting mutant sequence location and residue substitution as independent variables that imply the dependent disease phenotype and pathogenicity characteristics in 4 dimensional data points (4ddps). The SNP database contains entries with the majority having one or both dependent data entries unfulfilled. A neural network relating causes (mutant residue location and substitution) and effects (phenotype and pathogenicity) is trained, validated, and optimized using fulfilled 4ddps. It then predicts unfulfilled 4ddps providing the implicit disease model. A discrete Bayes network interprets fulfilled and predicted 4ddps with conditional probabilities for phenotype and pathogenicity given mutation location and residue substitution thus relating the neural network implicit model to explicit features of the motor and mybpc3 sequence and structural domains. Neural/Bayes network forecasting automates disease mechanism modeling by leveraging the world wide human missense SNP database that is in place and expanding.

KW - Autonomous motor

KW - Cardiac atrial myosin

KW - Cardiac myosin binding protein C

KW - Cardiac ventricular myosin

KW - Dilated cardiomyopathy

KW - Hypertrophic cardiomyopathy

KW - Inheritable heart disease mechanism

KW - Machine learning

KW - Restrictive cardiomyopathy

UR - http://www.scopus.com/inward/record.url?scp=85045540051&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85045540051&partnerID=8YFLogxK

U2 - 10.1016/j.yjmcc.2018.04.006

DO - 10.1016/j.yjmcc.2018.04.006

M3 - Article

C2 - 29654880

AN - SCOPUS:85045540051

VL - 119

SP - 19

EP - 27

JO - Journal of Molecular and Cellular Cardiology

JF - Journal of Molecular and Cellular Cardiology

SN - 0022-2828

ER -