1. Geuvadis RNAseq analysis @ UNIGE Genetic regulatory variants
Tuuli Lappalainen
University of Geneva
Geuvadis Analysis meeting II, July 11, 2012

2. Expression quantitative trait loci (eQTLs)
Expression level T
Genotypes
Works very well in cis. Difficult in trans
The same principle can be applied to any quantitative phenotype with a genomic locus
Statistical power only for common variants

3. QC – mRNA quantifications

4. miRNA QC
POP
LAB

5. eQTLs in Geuvadis Trans-analysis of large deletions didn’t yield much…
Pop N
Genes with eQTL (FDR)
Best eQTL indel (null 8.9%)
CEU+GBR
161
2608 (5.1%)
375 (14.4%)
TSI 92
1748 (7.7%)
242 (13.8%)
FIN 89
1822 (7.3%)
255 (14.0%)
YRI 77
2138 (6.3%)
242 (11.3%)
EUR union
342
3898 NA
ALL union
419
4895
Trans-analysis of large deletions didn’t yield much…
TODO:
Some methodological improvements
Combine Europeans with a PC correction of pop structure
Test exon versus transcript quantification

6. Splicing QTLs (sQTLs) in Geuvadis
Pop N
Genes with sQTL – transcript ratio
Genes with sQTL – links
CEU+GBR
161
121 (FDR 9.1%)
1251 forward (FDR 5.6%)
1077 reverse (FDR 6.6%)
nonredundant: 1949
ALL union
419
274 NA E1 E2 E3
FRE1-E2 = 5 (RE1-E2) / 5 (RE1-E2) + 3 (RE1-E3) = 0.625
FRE1-E3 = 3 (RE1-E3) / 5 (RE1-E2) + 3 (RE1-E3) = 0.375
links or junctions?
counts or fractions?
ALTRANS method by Halit Ongen

7. Integrating transcriptome QTLs
eQTLs for mRNA and miRNA
exon/miRNA_quantification ~ snp + covariates
sQTLs
link/junction_ratio ~ snp + covariates
link/junction quantification ~ snp + exon_quantification + covariates
multiple tQTLs: for the same gene
exon_quantification ~ snp2 + exon_eQTL_snp1 + covariates
link/junction ratio ~ snp2 + exon_eQTL_snp1 + covariates
targeted trans analysis
exon quantification ~ mi(eQTL)_snp + covariates
link/junction_ratio ~ mieQTL_snp + covariates

8. Functional annotation of eQTLs
TODO:
Direction of effect
TF motifs, PWM scores
Different eQTL frequencies
Other tQTLs
What’s the best way to tell if we have the causal variant or not? And how often do we seem to find it?

9. Allele specific expression
Statistical testing for ASE
What is the allelic ratio? Significantly different from 50-50?
cis eQTL*
coding SNP
mRNA-sequencing T T G T T T T A C C C
*or an epigenetic reason for higher expression of only one homolog in the studied cell population (e.g. imprinting)

10. Rare variants have higher effect sizes
ASE analysis
~ REGULATORY VARIANT FREQUENCY
derived allele frequency
power in eQTL analysis
eQTL analysis – expected result
Proper quantification of the effect?

11. Quantifying genetic effects to individual differences
TODO:
More work on the ASE difference analysis
Variation within/between populations
Rare variant ASE mapping

12. Can we predict functional effects of genetic variants?
How likely is an unknown variant to have regulatory effects based on known priors?
Gene expression ~ variant’s : distance from TSS + position in gene + functional annotation + allele frequency + conservation score + variant type…
“gene expression” could be e.g. exon quantification or link ratio
(Gaffney et al Genome Biology)
Does anyone have good experience of this type of modeling?

13. Acknowledgements
The FunPopGen lab Manolis Dermitzakis Analysis Alfonso Buil Thomas Giger Halit Ongen Data processing Ismael Padioleau Alisa Yurovsky Technicians Deborah Bielsen Emilie Falconnet Alexandra Planchon Luciana Romano Stanford School of Medicine Stephen Montgomery The 1000 Genomes Consortium Functional Interpretation Group FUNDING European Union National Institute of Health Louis-Jeantet Foundation Academy of Finland Emil Aaltonen Foundation Swiss National Science Foundation NCCR