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

Burghardt, Thomas P.; Ajtai, Katalin

doi:10.1016/j.yjmcc.2018.04.006

Cited by 19 publications

(20 citation statements)

References 43 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Pathogenicity is summarized as either pathogenic or benign by combining likely-pathogenic with pathogenic probabilities or likely-benign with benign probabilities (see Table 1 ). Demographic probabilities for each protein functional domain and pathogenicity category were computed with Bayesian statistics as described 34 and listed in SI Tables S7-S10 . An example from that data for a single functional domain in βmys is shown in Fig 5.…”

Section: Resultsmentioning

confidence: 99%

“…In vitro single myosin mechanical characterization uses purified and isolated myosin and reveals a telling correspondence between in vitro and in vivo systems that indicates myosin is to some extent an autonomous molecule such that the cardiac myosin is functionally the muscle in a molecule 47 . Autonomous myosin codes its mechanism for real time force-velocity regulation into the protein sequence that was captured in a Neural/Bayes network model 34 . Single residue sequence variation from a SNP in myosin or mybpc3 is a common cause of inheritable heart disease that affects people worldwide.…”

Section: Discussionmentioning

confidence: 99%

See 1 more Smart Citation

Demographic Model for Inheritable Cardiac Disease

Burghardt

2019

Preprint

Self Cite

View full text Add to dashboard Cite

19The cardiac muscle proteins, generating and regulating energy transduction during a heartbeat, 20 assemble in the sarcomere into a cyclical machine repetitively translating actin relative to myosin 21 filaments. Myosin is the motor transducing ATP free energy into actin movement against 22 resisting force. Cardiac myosin binding protein C (mybpc3) regulates shortening velocity 23 probably by transient N-terminus binding to actin while its C-terminus strongly binds the myosin 24 filament. Inheritable heart disease associated mutants frequently modify these proteins involving 25 them in disease mechanisms. Nonsynonymous single nucleotide polymorphisms (SNPs) cause 26 single residue substitutions with independent characteristics (sequence location, residue 27 substitution, human demographic, and allele frequency) hypothesized to decide dependent 28 phenotype and pathogenicity characteristics in a feed-forward Neural network model. Trial 29 models train and validate on a dynamic worldwide SNP database for cardiac muscle proteins 30 then predict phenotype and pathogenicity for any single residue substitution in myosin, mybpc3, 31 or actin. A separate Bayesian model formulates conditional probabilities for phenotype or 32 pathogenicity given independent SNP characteristics. Neural/Bayes forecasting tests SNP 33 pathogenicity vs (in)dependent SNP characteristics to assess individualized disease risk and in 34 particular to elucidate gender and human subpopulation bias in disease. Evident subpopulation 35 bias in myosin SNP pathogenicities imply myosin normally engages other sarcomere proteins 36 functionally. Consistent with this observation, mybpc3 forms a third actomyosin interaction 37competing with myosin essential light chain N-terminus suggesting a novel strain-dependent 38 mechanism adapting myosin force-velocity to load dynamics. The working models, and the 39 integral myosin/mybpc3 motor concept, portends the wider considerations involved in 40 understanding heart disease as a systemic maladaptation. 41 3 KEYWORDS 42 cardiac ventricular myosin, cardiac atrial myosin, cardiac myosin binding protein C, cardiac 43 actin, inheritable heart disease mechanism, machine learning, autonomous motor; hypertrophic 44 cardiomyopathy; dilated cardiomyopathy; restrictive cardiomyopathy; gender based risk 45 assessment 46 47 The cardiac muscle proteins generate and regulate energy transduction during a heartbeat. They 48 assemble into a cyclical machine in the sarcomere that repetitively translates actin relative to 49 myosin filaments. Myosin is the 50 motor transducing ATP free energy 51 to the work of moving actin against 52 resisting force. Cardiac myosin 53 binding protein C (mybpc3) 54 regulates shortening velocity 55 probably by binding transiently to 56 actin while stably bound to the 57 myosin filament.58Myosin (Fig 1) has a 140 59 kDa N-terminal globular head 60 called subfragment 1 (S1) and an 61 extended α-helical tail domain 62 (LMM+S2). Tail domains form 63 dimers that self-assemble into 64 myosin thick f...

show abstract

Section: Resultsmentioning

confidence: 99%

Section: Discussionmentioning

confidence: 99%

Demographic Model for Inheritable Cardiac Disease

Burghardt

2019

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…A consistent subset of the data trains and validates a feed-forward neural network modeling the contraction mechanism. The full database is completed using the model then interpreted probabilistically with a discrete Bayes network to provide the SNV probability for a functional domain location given pathogenicity and human population 19, 21 . Applying the approach to complex βmys/MYBPC3 identifies potential intra-protein pathways coupling these key proteins in the sarcomere pertaining to muscle shortening power generation and regulation in the human heart.…”

Section: Discussionmentioning

confidence: 99%

“…Protein domains (cd in Fig 3 panel a ) and their 2 letter abbreviations are indicated in Supplementary Information ( SI ) Table S1 . A protein complex made from four genes, MYH7, MYL2, MYL3, and MYBPC3 has domains (cd) from 65 functional sites combining assignments made previously 21 and new assignments in myosin including blocked head/converter interface (BH, index 3) 25 , mesa trail (MR, index 17) 26 , binding sites for C1, C2, and LT in MYBPC3 on myosin S2 (M1, M3, and M2, indices 18, 20, and 19) 27, 28 , for CX in MYBPC3 on myosin LM (M5, index 21) 29 , binding sites for titin and myomesin on myosin LM (T1 and Y1, indices 28 and 29) 29-31 , and for subdomains in RLC (RN, R1, RL, R2, and R3, indices 35-39) 32 . Every SNP in the database has an assigned domain.…”

Section: Methodsmentioning

confidence: 99%

“…A selected consistent subset of the data trains and validates a feed-forward neural network modeling the contraction mechanism. The database is completed using the model then interpreted probabilistically with a discrete Bayes network to give the SNV probability for a functional domain location given pathogenicity and human population as described earlier 19, 21 . Co-domains, intra-protein domains coupling βmys and MYBPC3, are identified by their population correlated SNV probability product for given pathogenicity.…”

mentioning

confidence: 99%

See 1 more Smart Citation

Natural variants with 2D correlation genetics identify domains coordinating sarcomere proteins during contraction

Burghardt

2021

Preprint

Self Cite

View full text Add to dashboard Cite

Muscle proteins assemble in a sarcomere then by coordinated action produce contraction force to shorten muscle. In the human heart ventriculum, cardiac myosin motor (βmys) repetitively converts ATP free energy into work. Cardiac myosin binding protein C (MYBPC3) in complex with βmys regulates contraction power generation. Their bimolecular complex βmys/MYBPC3 models the contractile system and is used here to study protein coupling. The database for single nucleotide variants (SNVs) in βmys and MYBPC3 surveys human populations worldwide. It consistently records SNV physical characteristics including substituted residue location in the protein functional domain, the side chain substitution, substitution frequency, and human population group, but inconsistently records SNV implicated phenotype and pathology outcomes. A selected consistent subset of the data trains and validates a feed-forward neural network modeling the contraction mechanism. The full database is completed using the model then interpreted probabilistically with a discrete Bayes network to give the SNV probability for a functional domain location given pathogenicity and human population. Co-domains, intra-protein domains coupling βmys and MYBPC3, are identified by their population correlated SNV probability product for given pathogenicity. Divergent genetics in human populations identify co-domain correlates in this method called 2D correlation genetics. Pathogenic and benign SNV data identify three critical regulatory sites, two in MYBPC3 with links to several domains across the βmys motor, and, one in βmys with links to the known MYBPC3 regulatory domain. Critical sites in MYBPC3 are hinges (one known another proposed) sterically enabling regulatory interactions with βmys. The critical site in βmys is the actin binding C-loop, a contact sensor triggering actin-activated myosin ATPase and contraction velocity modulator coordinating also with actin bound tropomyosin. C-loop and MYBPC3 regulatory domain linkage potentially impacts multiple functions across the contractile system. Identification of co-domains in a binary protein complex implies a capacity to estimate spatial proximity constraints for specific dynamic protein interactions in vivo opening another avenue for protein complex structure/function determination.

show abstract

Trust in Artificial Intelligence: Clinicians Are Essential

Bhatt

Shams

2021

Healthcare Information Technology for Cardiovascular Medicine

View full text Add to dashboard Cite

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

Cited by 19 publications

References 43 publications

Demographic Model for Inheritable Cardiac Disease

Demographic Model for Inheritable Cardiac Disease

Natural variants with 2D correlation genetics identify domains coordinating sarcomere proteins during contraction

Trust in Artificial Intelligence: Clinicians Are Essential

Contact Info

Product

Resources

About