Geuvadis RNAseq analysis @ UNIGE

Genetic regulatory variants

Tuuli Lappalainen

University of Geneva

Geuvadis Analysis meeting II, July 11, 2012

Expression quantitative trait loci (eQTLs)

T

C

Genotypes

Expre

ssio

n

level

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

QC – mRNA quantifications

miRNA QC

POP

LAB

eQTLs in Geuvadis

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 NA

TODO:Some methodological improvementsCombine Europeans with a PC correction of pop structureTest exon versus transcript quantification

Trans-analysis of large deletions didn’t yield much…

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

ALTRANS method by Halit Ongen

links or junctions?counts or fractions?

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

Functional annotation of eQTLs

TODO:Direction of effectTF motifs, PWM scoresDifferent eQTL frequenciesOther tQTLsWhat’s the best way to tell if we have the causal variant or not? And how often do we seem to find it?

Allele specific expression

A C

G T

CC

TT

T TT

cis eQTL* coding SNP mRNA-sequencingStatistical testing for ASE

What is the allelic ratio? Significantly different from 50-50?

*or an epigenetic reason for higher expression of only one homolog in the studied cell population (e.g. imprinting)

Rare variants have higher effect sizes

0 0.5 1 derived allele frequency

pow

er

in e

QTL

analy

sis

eQTL analysis – expected result

ASE analysis

~ REGULATORY VARIANT FREQUENCY

Proper quantification of the effect?

Quantifying genetic effects to individual differences

TODO:More work on the ASE difference analysisVariation within/between populationsRare variant ASE mapping

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. 2012 Genome Biology)

Does anyone have good experience of this type of modeling?

Acknowledgements

The FunPopGen lab

Manolis Dermitzakis

AnalysisAlfonso BuilThomas GigerHalit Ongen

Data processingIsmael PadioleauAlisa Yurovsky

TechniciansDeborah BielsenEmilie FalconnetAlexandra PlanchonLuciana Romano

Stanford School of MedicineStephen Montgomery

The 1000 Genomes ConsortiumFunctional Interpretation Group

FUNDINGEuropean UnionNational Institute of HealthLouis-Jeantet FoundationAcademy of FinlandEmil Aaltonen FoundationSwiss National Science FoundationNCCR