- Количество слайдов: 18
Challenges of an Epidemiologist Working in Genomics Wendy Post, MD, MS Associate Professor of Medicine and Epidemiology Cardiology Division Johns Hopkins University
“There is a need to bridge the chasm between geneticists and traditional epidemiologists who are now wondering how they can apply GWAS technology to their studies”. Teri Manolio 5/8/07
Nature Genetics 2006; 38(6): 644 -51 (epub Apr 30 2006).
CAPON Association with adjusted QT interval Results of a genome wide association study in KORA S 4 and 2 replication cohorts Cohort N Effect MAF Adjusted p KORA S 4 3366 4. 9 msec 36% < 10 -7 KORA F 3 2646 7. 9 msec 36% < 10 -11 FHS 1805 4. 0 msec 39% 0. 004 *QT- adjusted for age, gender and heart rate Arking DE, Pfeufer A, Post W n al. Nature Genetics; published online Apr 30 2006. et
Heritability of Left Ventricular Mass The Framingham Heart Study Wendy S. Post; Martin G. Larson; Richard H. Myers; Maurizio Galderisi; Daniel Levy Hypertension. 1997; 30: 1025 -1028. © 1997 American Heart Association, Inc. From the National Heart, Lung, and Blood Institute's Framingham Heart Study, Framingham, Mass (W. S. P. , M. G. L. , R. H. M. , M. G. , D. L. ); the Division of Cardiology, Beth Israel Hospital, Boston, Mass (D. L. ); the Department of Neurology (R. H. M. ), Division of Epidemiology and Preventive Medicine (M. G. L. , D. L. ), Boston University School of Medicine; the National Heart, Lung, and Blood Institute, Bethesda, Md (D. L. ), University of Naples, Italy (M. G. ); and the Division of Cardiology, Johns Hopkins Hospital, Baltimore, Md (W. S. P. ).
Confusing genetics nomenclature • Before rs numbers the snp names kept changing – Makes it hard to compare results to prior studies in Pub. Med – rs numbers (Ref. SNP accession ID- db SNP) • db SNP- reference database (www. dbsnp. com) • Forward strand versus reverse strand – Nucleotide names • Dominant model versus recessive model – Relative to major or minor allele? • AB+AB vs BB • Remembering my biochemistry – Untranslated exon? • Exon= region of DNA transcribed into the final m. RNA
Complicated authorship issues • Collaboration is key – – Phenotypers Statisticians Bioinformatics Genotypers • Collaboration with other cohorts for replication/validation • Order of authorship on manuscripts is not straightforward – Decide before the work is done
What covariates to put in the model? • Epidemiologists “worry” a lot about confounding. • Confounders are associated with the outcome (phenotype) and the predictor (genotype). – most of our traditional confounders are not associated with genotype. • Might want to add covariates for “precision” • How much of the variability in the phenotype is explained by genotype after including known predictors in the model?
Choosing appropriate control groups Epidemiology 101 Cases and controls need to be collected in a similar fashion similar ancestry similar environmental exposures
Dealing with population stratification • How big of an issue is it really? • Should we use AIMs or self described race/ethnicity? – AIM (ancestral informative markers) • allele frequencies of snps differ based on parental population • Can estimate the ancestral proportion of an individual – Self described race/ethnicity • When can we combine racial/ethnic groups for analyses when there is no statistical interaction?
Gene-environment and gene interactions • Complex disorders – Multiple genes and environmental interactions • Tests for interactions – Multiple testing issues – Power • How to combine multiple genes/snps into same prediction model
Multiple testing issues • Fishing expedition – Traditionally in epidemiology, seen as “poor science” – GWAS is a really big, sophisticated, fishing expedition • Fishing in Alaska for seven different kinds of salmon, instead of fishing on the LI sound.
What p value do we use? • Bonferroni adjustment seems overly conservative – False negatives • False Discovery Rate • Need for replication/validation – What cutpoint do we use to move results forward?
Other issues • Lack of reproducibility – False positives versus • differences in environmental exposures or haplotype structure • different study design • HWE – Relative frequency of alleles for a snp are stable in the population (not changing over successive generations). • p 2, 2 pq, q 2 • What genetic model to test – 2 df, additive, dominant, recessive • Again, issues of multiple testing arise
To patent or not to patent our results • Epidemiologists rarely patent findings • History of new scientific discoveries in genetics acquiring patents • Could hinder scientific progress?
Ann. Int. Med. 49: 556 -567, 1958