A quantitative estimate of individual genetic vulnerability for a multifactorial condition

Application to bipolar illness

John Nurnberger, Tatiana Foroud, E. T. Meyer, K. L. Hu, L. Flury, J. Su

Research output: Contribution to journalArticle

Abstract

We have developed an algorithm that may be applied to genomic survey data to derive estimates of genetic vulnerability for each person in a pedigree. The method utilizes data from GENEHUNTER to identify loci of interest. A 'sharing score' is computed for each allele in the pedigree at a given locus. An individual's 'locus vulnerability score' is the sum of scores for each of his or her alleles multiplied by the NPL score at the locus. The total 'vulnerability score' is then corrected for the genotypic information available for the individual at the loci of interest. This method has been applied to the NIMH Genetics Initiative Bipolar initial dataset (N = 539 persons). We have used the output from a GENEHUNTER analysis using a diagnostic model in which SA/BP, BPI, BPII, and UPR are affected, while all others are unaffected. The 33 loci with Z scores > 1.0 were included in the analysis. A bimodal distribution of vulnerability scores is produced, with affected individuals generating higher scores than unaffected (this is to be expected from the algorithm). We have also observed an effect of diagnostic subtype, such that BPI=SA/BP > BPII=UPR (Kruskal-Wallis chi square = 15.8, p= .0012). We predicted that early onset probands would have higher vulnerability scores than late onset (p=.03, controlling for the interaction of diagnosis and age of onset). These results are consistent with the assumption of multifactorial inheritance and with the hypothesis that some of the 33 loci tested are true genetic signals. They also indicate that the algorithm, while necessarily employing a crude approximation to the underlying genetics of bipolar illness, is nevertheless capable of producing clinically meaningful results in this dataset. These methods might be used to attempt prediction of illness in at-risk individuals as a part of research on risk and protective factors. In the present data, scaled vulnerability scores for unaffected persons range from 27 to 68; scores for affected individuals range from 31 to 95. Persons with a score of >55 are more likely to be affected than unaffected; those with a score of < 55 are more likely to be unaffected than affected. Using this cut-off point both sensitivity and specificity are 77%. However, it would not be appropriate for 'vulnerability scores' to be used in any form of clinical genetic counseling at this point in time. The method described is sufficiently general that it may be useful in analyzing genomic survey data from other conditions with complex inheritance. It is easily modifiable to incorporate interactions between loci as they become known or suspected. A computerized version will be available.

Original languageEnglish
Pages (from-to)487
Number of pages1
JournalAmerican Journal of Medical Genetics, Part B: Neuropsychiatric Genetics
Volume96
Issue number4
StatePublished - Aug 7 2000

Fingerprint

Pedigree
Alleles
Multifactorial Inheritance
National Institute of Mental Health (U.S.)
Genetic Counseling
Age of Onset
Sensitivity and Specificity
Research
Surveys and Questionnaires
Datasets
Protective Factors

ASJC Scopus subject areas

  • Genetics(clinical)
  • Neuropsychology and Physiological Psychology
  • Neuroscience(all)

Cite this

@article{58d4b0744416414d8d70992baf9d7b51,
title = "A quantitative estimate of individual genetic vulnerability for a multifactorial condition: Application to bipolar illness",
abstract = "We have developed an algorithm that may be applied to genomic survey data to derive estimates of genetic vulnerability for each person in a pedigree. The method utilizes data from GENEHUNTER to identify loci of interest. A 'sharing score' is computed for each allele in the pedigree at a given locus. An individual's 'locus vulnerability score' is the sum of scores for each of his or her alleles multiplied by the NPL score at the locus. The total 'vulnerability score' is then corrected for the genotypic information available for the individual at the loci of interest. This method has been applied to the NIMH Genetics Initiative Bipolar initial dataset (N = 539 persons). We have used the output from a GENEHUNTER analysis using a diagnostic model in which SA/BP, BPI, BPII, and UPR are affected, while all others are unaffected. The 33 loci with Z scores > 1.0 were included in the analysis. A bimodal distribution of vulnerability scores is produced, with affected individuals generating higher scores than unaffected (this is to be expected from the algorithm). We have also observed an effect of diagnostic subtype, such that BPI=SA/BP > BPII=UPR (Kruskal-Wallis chi square = 15.8, p= .0012). We predicted that early onset probands would have higher vulnerability scores than late onset (p=.03, controlling for the interaction of diagnosis and age of onset). These results are consistent with the assumption of multifactorial inheritance and with the hypothesis that some of the 33 loci tested are true genetic signals. They also indicate that the algorithm, while necessarily employing a crude approximation to the underlying genetics of bipolar illness, is nevertheless capable of producing clinically meaningful results in this dataset. These methods might be used to attempt prediction of illness in at-risk individuals as a part of research on risk and protective factors. In the present data, scaled vulnerability scores for unaffected persons range from 27 to 68; scores for affected individuals range from 31 to 95. Persons with a score of >55 are more likely to be affected than unaffected; those with a score of < 55 are more likely to be unaffected than affected. Using this cut-off point both sensitivity and specificity are 77{\%}. However, it would not be appropriate for 'vulnerability scores' to be used in any form of clinical genetic counseling at this point in time. The method described is sufficiently general that it may be useful in analyzing genomic survey data from other conditions with complex inheritance. It is easily modifiable to incorporate interactions between loci as they become known or suspected. A computerized version will be available.",
author = "John Nurnberger and Tatiana Foroud and Meyer, {E. T.} and Hu, {K. L.} and L. Flury and J. Su",
year = "2000",
month = "8",
day = "7",
language = "English",
volume = "96",
pages = "487",
journal = "American Journal of Medical Genetics, Part B: Neuropsychiatric Genetics",
issn = "1552-4841",
publisher = "Wiley-Liss Inc.",
number = "4",

}

TY - JOUR

T1 - A quantitative estimate of individual genetic vulnerability for a multifactorial condition

T2 - Application to bipolar illness

AU - Nurnberger, John

AU - Foroud, Tatiana

AU - Meyer, E. T.

AU - Hu, K. L.

AU - Flury, L.

AU - Su, J.

PY - 2000/8/7

Y1 - 2000/8/7

N2 - We have developed an algorithm that may be applied to genomic survey data to derive estimates of genetic vulnerability for each person in a pedigree. The method utilizes data from GENEHUNTER to identify loci of interest. A 'sharing score' is computed for each allele in the pedigree at a given locus. An individual's 'locus vulnerability score' is the sum of scores for each of his or her alleles multiplied by the NPL score at the locus. The total 'vulnerability score' is then corrected for the genotypic information available for the individual at the loci of interest. This method has been applied to the NIMH Genetics Initiative Bipolar initial dataset (N = 539 persons). We have used the output from a GENEHUNTER analysis using a diagnostic model in which SA/BP, BPI, BPII, and UPR are affected, while all others are unaffected. The 33 loci with Z scores > 1.0 were included in the analysis. A bimodal distribution of vulnerability scores is produced, with affected individuals generating higher scores than unaffected (this is to be expected from the algorithm). We have also observed an effect of diagnostic subtype, such that BPI=SA/BP > BPII=UPR (Kruskal-Wallis chi square = 15.8, p= .0012). We predicted that early onset probands would have higher vulnerability scores than late onset (p=.03, controlling for the interaction of diagnosis and age of onset). These results are consistent with the assumption of multifactorial inheritance and with the hypothesis that some of the 33 loci tested are true genetic signals. They also indicate that the algorithm, while necessarily employing a crude approximation to the underlying genetics of bipolar illness, is nevertheless capable of producing clinically meaningful results in this dataset. These methods might be used to attempt prediction of illness in at-risk individuals as a part of research on risk and protective factors. In the present data, scaled vulnerability scores for unaffected persons range from 27 to 68; scores for affected individuals range from 31 to 95. Persons with a score of >55 are more likely to be affected than unaffected; those with a score of < 55 are more likely to be unaffected than affected. Using this cut-off point both sensitivity and specificity are 77%. However, it would not be appropriate for 'vulnerability scores' to be used in any form of clinical genetic counseling at this point in time. The method described is sufficiently general that it may be useful in analyzing genomic survey data from other conditions with complex inheritance. It is easily modifiable to incorporate interactions between loci as they become known or suspected. A computerized version will be available.

AB - We have developed an algorithm that may be applied to genomic survey data to derive estimates of genetic vulnerability for each person in a pedigree. The method utilizes data from GENEHUNTER to identify loci of interest. A 'sharing score' is computed for each allele in the pedigree at a given locus. An individual's 'locus vulnerability score' is the sum of scores for each of his or her alleles multiplied by the NPL score at the locus. The total 'vulnerability score' is then corrected for the genotypic information available for the individual at the loci of interest. This method has been applied to the NIMH Genetics Initiative Bipolar initial dataset (N = 539 persons). We have used the output from a GENEHUNTER analysis using a diagnostic model in which SA/BP, BPI, BPII, and UPR are affected, while all others are unaffected. The 33 loci with Z scores > 1.0 were included in the analysis. A bimodal distribution of vulnerability scores is produced, with affected individuals generating higher scores than unaffected (this is to be expected from the algorithm). We have also observed an effect of diagnostic subtype, such that BPI=SA/BP > BPII=UPR (Kruskal-Wallis chi square = 15.8, p= .0012). We predicted that early onset probands would have higher vulnerability scores than late onset (p=.03, controlling for the interaction of diagnosis and age of onset). These results are consistent with the assumption of multifactorial inheritance and with the hypothesis that some of the 33 loci tested are true genetic signals. They also indicate that the algorithm, while necessarily employing a crude approximation to the underlying genetics of bipolar illness, is nevertheless capable of producing clinically meaningful results in this dataset. These methods might be used to attempt prediction of illness in at-risk individuals as a part of research on risk and protective factors. In the present data, scaled vulnerability scores for unaffected persons range from 27 to 68; scores for affected individuals range from 31 to 95. Persons with a score of >55 are more likely to be affected than unaffected; those with a score of < 55 are more likely to be unaffected than affected. Using this cut-off point both sensitivity and specificity are 77%. However, it would not be appropriate for 'vulnerability scores' to be used in any form of clinical genetic counseling at this point in time. The method described is sufficiently general that it may be useful in analyzing genomic survey data from other conditions with complex inheritance. It is easily modifiable to incorporate interactions between loci as they become known or suspected. A computerized version will be available.

UR - http://www.scopus.com/inward/record.url?scp=33749083491&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=33749083491&partnerID=8YFLogxK

M3 - Article

VL - 96

SP - 487

JO - American Journal of Medical Genetics, Part B: Neuropsychiatric Genetics

JF - American Journal of Medical Genetics, Part B: Neuropsychiatric Genetics

SN - 1552-4841

IS - 4

ER -