Lecture 9+10: Drug discovery I Flashcards
Purpose of PGx research (not starred)
-test genotype/phenotype correlation
-explain pt dif in phenotype
-find alleles that affect phenotype
-establish relationship between variation and phenotype
-apply
Case Control Study
-Collect genotype of case and control populations (pt w virus clearance vs persistance ex)
-table of genotypes
-convert number of genotypes into number of alleles = new table
Contingency table (not starred)
-chi-square or fisher’s exact test
-null hypothesis: NO association between drug and allele
-small X2 = support hypothesis (no association)
-large X2= reject null hypothesis (association)
-P-value
Interpretation of P value
- P > 0.1: NO association (accept null)
- 0.05 < P < 0.1: marginal association
- P < 0.05: significant association (reject null)
- P < 0.01: very strong assocation (reject null)
P-value notes
-does NOT measure stength of association relationship
-can be affected by sample size (bigger size = lower p value even under same freq)
-can be affected by allele frequency
Measure of Strength
-Odds ratio
-Hazard ratio (mostly survival data)
Odds ratio
-inc risk for phenotype by carrying specific genotype/allele compared to the patients without carrying
=odds of phenotype in individual w allele divided by odds of phenotype in individual w/o genotype/allele
Calculation of OR example
- (T allele in persistant / T allele in cleared) / (C allele persistant / C allele cleared)
-(539/198) / (701/578) = 2.24
-pt w T allele have 2.24 times more of a chance to develop persistance compared to pt w C allele
OR interpretation
-OR < 1: potential dec risk (protective allele) (smaller the OR = lower risk)
-OR = 1: NO association
-OR > 1: potential inc risk (risk allele) (higher OR = higher risk)
How do we know if OR is significant or just chance?
-95% Confidence interval
95% CI
-over 95% of probability that association is confident
-statistical probability for OR
95% CI interpretation
-CI < 1: significant PROTECTIVE effect
-CI contains 1: NO significance
-CI > 1: significant RISK
PGx study design (not starred) How do we identify SNP associated w phenotype?
-example only involves 1 SNP
-candidate gene/SNP approach: hypothesis based design (best guess)
-Genome-wide association study (GWAS): don’t need hypothesis to test, less common, use -log(p)
Correction for P values
-high probability of many SNPs associated w phenotype (false positive), so called multiple-testing
-more SNPs tested = higher probability of false positive
-bonferroni correction
Bonferroni correction
-corrected P=0.05/N (total number of SNPs tested)
-in general, use 5 x 10^-8 as corrected significant GWAS P-value
Experiment design
-control is key (positive or negative)
-know distribution of data
-replication important
Control
-positive (standard of care drug)
-negative (sham drug, unethical in human trials)
-experiment group (investigational drug)
-need large sample size for reliable results
-know range of data (upper and lower limit)
Distribution of data in experiment
-median preferred over mean bc faster and patient data may not be normally distributed
Replication of experiment (not starred)
-may not be representative of all populations
-even if p value is very low, finding can still be chance
-association study usually requires independent replications in other sample sets to increase n number (sample size is v important)
Correlation not casual
-not always cause-effect
Question
none of the above
