Transcript International Collaboration for the Genomics of HIV

IAS Workshop
2 July 2013
Combining –omics to study
the host and the virus
Jacques Fellay
School of Life Sciences
École Polytechnique Fédérale de Lausanne - EPFL
Lausanne, Switzerland
Allele frequency of variant
<<<<<1%
>5%
Sequencing
studies
+++
How to look for
associations with
DNA variants?
++
Clinical impact
Genome-wide
association studies
+
HIV host genetic studies:
clinical phenotypes
Viral control disease progression
Resistance acquisition
Exposure
Infection
Science 2007 Aug 17;317(5840):944-7
Science 2010 Dec 10;330(6010):1551-7
Science 1996 Sep 27;273(5283):1856-62
Where do we go from here?
1. More common variants?

Meta-analysis of GWAS data
2. Rare functional variants?

Sequencing
3. Host impact on viral sequence?

“Genome-to-genome” interaction analysis
More common variants?
International Collaboration
for the Genomics of HIV
Objective: combine existing GWAS data from HIV+
cohorts to conduct joint analyses:
- of viral control and/or disease progression
- of HIV susceptibility:
After QC and imputation, comparison between
6300 HIV infected cases and 7300 population
controls over 5x106 variants
B*57:01
B*27:05
Frailty bias
Due to their shorter survival time, patients with rapid disease
progression are underrepresented in “chronic” cohorts, while
individuals with prolonged disease-free survival times are more
likely to be included
Analysis restricted to patients with known date of infection
rs4418214 p=0.01
International Collaboration
for the Genomics of HIV
 HIV acquisition: no significant associations
(after accounting for survivor bias), with the
exception of CCR5Δ32 homozygosity: p=3E13. No replication of all other previously
reported associations (N=22)
 McLaren et al., PLoS Pathogens, in press
 HIV control: analyses are ongoing
Rare functional variants?
 Polymorphisms of strong effect are kept at low
frequency by evolutionary forces
 Rare, functional variants are not well
represented by GWAS
 Sequencing has been
highly successful for
uncovering causes of
rare Mendelian diseases
Patient sample
N=400
Target enrichment
Variant calling and
frequency estimation
www.broadinstitute.org/gatk
DNA extraction and
quantity normalization
Sequencing
(paired end reads)
Variant annotation
CAA GTA AAC ATA GGA CTT CTT
CAA GTA AAC ATA GGA CAT CTT
snpeff.sourceforge.net
DNA pooling and
bar-coding
Association testing
with HIV VL
Alignment and base
quality recalibration
T/C
Single
variant
Gene burden
Exome sequencing performance
Metric
Mean coverage
% Covered >5x
Call rate
GWAS concordance
Per sample
Total non-ref
Non-synonymous
Loss of function
Ti/Tv
Score
73x
94.0%
99.9%
99.0%
Score
16,105
8,122
39
3.21
Single variant results (MAF > 1%)
MHC signal consistent with GWAS
Can be explained by variation in HLA-B (B*57:01) and HLA-C (3’ UTR)
Single variant results (MAF > 1%)
No single variant associates with spVL after accounting for known signals
Burden testing
• Gene-based (~20,000 tests)
• Set-based
siRNA
Screens
Interacting
Proteins
Burden testing
 HIV-specific sets from the literature
 I HIV dependency factors
 II HIV/Human PPI by MS
 III Interferon stimulated
genes
 IV HIV interactome
No significant
 Union set = associations
2,791
 Intersection (2 or more) = 292
 Restrict analysis to nonsynonymous and loss of
function variants
Host genomics of HIV disease:
Limitations of clinical phenotypes
1. Good phenotypes are hard to get:
- Long follow-up of patients
- Close collaboration with clinicians
- It’s now unethical to observe the natural
history of HIV infection
2. Clinical outcomes are quite far from
potentially causal gene variants
Host genomics
Host-pathogen genomics
The principle of Genome-to-Genome analysis
Escape
mutations
Genetic variants
Host restriction factors leading to viral escape can be
uncovered by searching for their imprints on viral genomes
HIV-1 “genome-to-genome”
study
• 1100 study
participants
• Caucasians
infected with
subtype B HIV-1
• Paired genetic data:
 Human: genome-wide genotypes from GWAS
 HIV-1: full-length consensus sequence
3 sets of genome-wide
comparisons
Human genetic
variation
1 GWAS
Viral load
2077 GWAS
(1 per variable HIV
amino acid
present in >20
samples)
1 proteome-wide
association
study
(2077 linear
regressions)
HIV-1 amino
acid variants
Human
SNPs
Viral Load
HIV
sequence
mutations
Human
SNPs
Viral Load
HIV
sequence
mutations
SNPs, HLA and CTL epitopes
Association of HIV-1 amino acids with
VL
Human
SNPs
Viral Load
HIV
sequence
mutations
No significant
association
Changes in VL for amino acid variants
associated with rs2395029 / B*57:01
(p<0.001)
Conclusions
• Using viral variation as an intermediate
phenotype can be a sensitive method for
detecting host associations
• Can be applied to other infectious diseases
HIV host genetics – the way forward
1. More samples
– Host genetics of infectious disease outcome still
lags far behind other complex traits in terms of
power
2. More variants
– Current technologies still do not provide a
complete picture of human genetic variation
3. More phenotypes
– Easily measured, intermediate phenotypes can
provide a potentially powerful method for
detection of important loci
Paul McLaren
Istvan Bartha
Thomas Junier
Samira Asgari
Ana Bittencourt
All ICGH collaborators
University of Lausanne
Microsoft Research
Duke University
Genomic Technologies Facility
Vital-IT Computing Center
Amalio Telenti
David Heckerman
David Goldstein
Keith Harshman
Ioannis Xenarios