development of diagnostic enzyme assay to detect leuser virus
DFEnitely Different: The joint distribution of mutation fitness effects between populations
1. Ryan Gutenkunst
Dept of Molecular and Cellular Biology
University of Arizona
DFEnitely Different
The joint distribution of mutation fitness
effects between populations
http://gutengroup.mcb.arizona.edu
with
Alyssa Fortier,Alec Coffman,Travis Struck,
Jose Burguete,Aaron Ragsdale
@RyanGutenkunst
pop1 pop2
S
2
S1S1
2. Distribution of fitness effects (DFE)
of new mutations
Frequency/density
SelecƟon coefficient
(B)
0
Trends in Genetics December 2014, Vol
Frequency/density
Frequency/density
Beneficial
EffecƟvely neutral
Deleterious
Strongly deleterious/lethal
SelecƟon coefficient
0
(A) (B)
Bank et al. (2014) Trends Genet
5. p1 = 0.0
p1 = 0.5
p1 = 1.0
pop1 pop2
S
2
S1S1
Joint Distribution of Fitness Effects
• GxE and GxG interactions
• Early stages of speciation
• GWAS transferability
ρ=1
lognorm
al
w
eightp1
ρ=0 lognormal
weight 1-p1
6. Figure S6: Inferred values of p1 for human functional groups. Whiskers deno
Inference
• Infer model of demographic history from synonymous
SNPs (using dadi)
• Cache expected frequency spectra for grid of s1, s2
(code built off fitdadi, Kim et al. (2017) Genetics)
• Expected spectrum for a
given DFE is a weighted sum
over that cache
• Optimize parameters of DFE
by maximizing likelihood of
nonsynonymous SNP data,
as Poisson Random Field
8. Model misspecification
Figure 2: Power and robustness of joint DFE inference. Using sim
power of joint DFE inference and its robustness to model misspecifica
mixture proportion p1 for the perfectly correlated component. A: Preci
for small sample sizes. What was ✓ for these simulations? How many
✓s was assumed to be 5537. The average SNP counts for the sets of sa
size 2, 10871 for sample size 4, 13107 for sample size 6, 22308 for samp
and 44897 for sample size 100. B: Precise inference of p1 was possible e
recently diverged. Should extend this to T = 10 4
, 10 5
, to see failure b
values of p1 was modest when data were simulated with a demographi
simulated data, we tested the statistical
ification, focusing on inferences of the
Demographic history
Figure 2: Power and robustness of joint DFE inference. Using simulated da
Dominance
Figure 2: Power and robustness of joint DFE inference. Using simula
power of joint DFE inference and its robustness to model misspecificatio
mixture proportion p1 for the perfectly correlated component. A: Precise
for small sample sizes. What was ✓ for these simulations? How many SN
✓s was assumed to be 5537. The average SNP counts for the sets of samp
size 2, 10871 for sample size 4, 13107 for sample size 6, 22308 for sample s
and 44897 for sample size 100. B: Precise inference of p1 was possible even
recently diverged. Should extend this to T = 10 4
, 10 5
, to see failure befo
values of p1 was modest when data were simulated with a demographic m
growth and migration, but analyzed assuming no growth or no migration. D
Figure 2: Power and robustness of jo
power of joint DFE inference and its r
mixture proportion p1 for the perfectly
for small sample sizes. What was ✓ for
✓s was assumed to be 5537. The averag
size 2, 10871 for sample size 4, 13107 fo
and 44897 for sample size 100. B: Precis
recently diverged. Should extend this to
values of p1 was modest when data wer
growth and migration, but analyzed assu
Gamma vs. lognormal DFE
ness of joint DFE inference. Using simulated data, we tested the statistical
Mixture vs. bivariate
lognormal DFE
9. Drosophila application
Pool et al. (2015)
Genetics
Joint DFE
from nonsynonymous SNPs
p1 = 0.967 ± 0.022
Demographic history
from synonymous SNPs
ZambiaFrance
15. Summary
• The joint DFE between populations can be effectively
inferred from the joint allele frequency spectrum
• In Drosophila, fitness effects tend to be highly
correlated between populations, but this correlation
varies with gene function
• The fitness effects of more deleterious mutations are
less correlated between populations
• The correlation between fitness effects is higher in
humans than Drosophila
• Very interested in collaborating to infer the joint DFE
in other systems
@RyanGutenkunst
http://gutengroup.mcb.arizona.edu