forked from dyerlab/applied_population_genetics
-
Notifications
You must be signed in to change notification settings - Fork 0
/
parentage.rmd
568 lines (345 loc) · 14.8 KB
/
parentage.rmd
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
# Parentage Analyses {.imageChapter}
<div class="chapter_image"><img src="chapter_images/ch_butterfly.jpg"></div>
```{r echo=FALSE, message=FALSE, warning=FALSE}
library(knitr)
library(ggplot2)
library(gstudio)
```
The analysis of parental and offspring data
1. Parentage-type analyses.
2. Analyses based upon multiple paternity
Parentage analyses are used in a broad range of studies:
* Identical vs fraternal twins
* Agricultural crop line differentiation
* Differentiate between livestock/dog/cat breeds
* Pathogenic strain identification (e.g., Hep-C strains *a-e*)
* Assigning parentage to individuals
* Assigning individuals to populations
* Identifying the source of unknown tissues (e.g., gettin' perps)
```{r echo=FALSE}
knitr::include_graphics("media/paternity.png")
```
A statistical approach for identifying the parent(s) of a particular individual. This requires:
1. A set of genetic markers that are bi-parentally inherited
2. Variation in these markers
3. Some assumptions about the *prior* probability of the union of parents.
## Paternity vs. Maternity
For single parent parentage analysis, it is either paternity or maternity that is being established. Here we assume that the other parent is definitely the biological parent of the individual (e.g., $P(prior)=1$). This can be because:
1. The offspring was collected from the identified parent
2. There is other evidence that points to the identified parent being the biological one.
The goal then is to determine who the unidentified parent is with some level of statistical probability.
Terms Used in Parentage
The following terms are commonly used in parentage analyses:
**Extra-Pair Paternity** - Fertilization resulting from copulation outside a recognized pair bond.
**Multiple Paternity** - Offspring produced from mating events with different sets of individuals.
**Paternity/Maternity Exclusion** - Excluding an individual based upon an in-congruence in observed genetic data.
## Probability of Exclusion
Mother | Offspring | Excluded Dads | Probability
--------------------|---------------------------|---------------------|--------------
$A_1A_1\;(p^2_1)$ | $A_1A_1\;(p_1)$ | $A_2A_2\;(p_2^2)$ | $p_1^3p_2^2$
$A_1A_1\;(p^2_1)$ | $A_1A_2\;(p_2)$ | $A_1A_1\;(p_1^2)$ | $p_1^4p_2$
$A_1A_1\;(p^2_1)$ | $A_2A_2\;(0)$ | - | -
$A_1A_2\;(2p_1p_2)$ | $A_1A_1\;(\frac{p_1}{2})$ | $A_2A_2\;(p_2^2)$ | $p_1^2p_2^3$
$A_1A_2\;(2p_1p_2)$ | $A_1A_2\;(\frac{1}{2})$ | - | -
$A_1A_2\;(2p_1p_2)$ | $A_2A_2\;(\frac{p_2}{2})$ | $A_1A_1\;(p_1^2)$ | $p_1^3p_2^2$
$A_2A_2\;(p^2_2)$ | $A_1A_1\;(0)$ | - | -
$A_2A_2\;(p^2_2)$ | $A_1A_2\;(p_1)$ | $A_2A_2\;(p_2^2)$ | $p_1p_2^4$
$A_2A_2\;(p^2_2)$ | $A_2A_2\;(p_2)$ | $A_1A_1\;(p_1^2)$ | $p_1^2p_2^3$
Probability of Exclusion
\[
P_{exc} = p_1^3p_2^2 + p_1^4p_2 +p_1^2p_2^3 +p_1^3p_2^2 +p_1p_2^4 +p_1^2p_2^3 \\
\]
which when simplified down a bit becomes
\[
P_{exc} = p_1p_2(1-p_1p_2)
\]
Single Locus Paternity Exclusion
```{r chunk-parentage}
p <- seq(0,1,by=0.02)
q <- 1-p
Pexcl <- p*q*(1-p*q)
plot(Pexcl ~ p, xlab="Allele frequency, p", ylab="Paternity Exclusion Probability")
```
Multilocus Exclusion
Exclusion probabilities are multiplicative properties.
\[
P_{excl} = 1 - \prod_{i=1}^\ell(1-P_{excl,\;i})
\]
Example
```{r}
p <- rep( 0.5*0.5*(1-0.5*0.5), 5 )
p
ptot <- 1
for( i in 1:length(p))
ptot <- ptot * (1-p[i])
1-ptot
```
Mother | Offspring | Excluded Father $A_xA_y$ | Probability of Exclusion
--------------------|-------------------------------|---------------------------------|------------------------------------
$A_iA_i\;(p_i^2)$ | $A_iA_i\;(p_i)$ | $x,y\ne i\;\; (1-p_i)^2$ | $p_i^3(1-p_i)^2$
$A_iA_i\;(p_i^2)$ | $A_iA_j\;(p_j)$ | $x,y\ne j\;\; (1-p_j)^2$ | $p_i^2p_j(1-p_i)^2$
$A_iA_j\;(2p_ip_j)$ | $A_iA_i\;(\frac{p_i}{2})$ | $x,y \ne i\;\; (1-p_i)^2$ | $p_i^2p_j(1-p_i)^2$
$A_iA_j\;(2p_ip_j)$ | $A_iA_j\;(\frac{p_i+p_j}{2})$ | $x,y \ne i,j\;\; (1-p_i-p_j)^2$ | $p_ip_j(p_i+p_j)(1-p_i-p_j)^2$
$A_iA_j\;(2p_ip_j)$ | $A_iA_k\;(\frac{p_k}{2})$ | $x,y \ne k\;\; (1-p_k)^2$ | $p_ip_jp_k(1-p_k)^2$
$A_iA_j\;(2p_ip_j)$ | $A_jA_k\;(\frac{p_k}{2})$ | $x,y \ne k\;\; (1-p_k)^2$ | $p_ip_jp_k(1-p_k)^2$
$A_iA_j\;(2p_ip_j)$ | $A_jA_j\;(\frac{p_j}{2})$ | $x,y \ne j\;\; (1-p_j)^2$ | $p_ip_j^2(1-p_j)^2$
## Paternity Exclusion
Likelihood Ratios
A likelihood ratio is given by:
\[
LR = \frac{H_P}{H_D}
\]
where the $H_X$ values are the hypotheses probabilities.
Nomenclature For Parentage
Individual | Identifier | Genotype
---------------------|--------------|------------
Female Parent | $FP_i$ | $\alpha_i$
Putative Male Parent | $MP_j$ | $\beta_j$
Offspring | $O_k$ | $\gamma_k$
$\;$
**Paternal Probability** The posterior odds of paternity versus non-paternity given the totality of genetic information.
Likelihood Ratios | Genetic Equivalences
The likelihood of one hypothesis, $H_1$ relative to another $H_2$ is:
\[
L(H_1,H_2|D) = \frac{P(D|H_1)}{P(D|H_2)}
\]
where
\[
P(D|H) = T(\gamma | \alpha, \beta)P(\alpha)P(\beta)
\]
Assuming $H_1:$ states that $\beta$ is the real father of $\gamma$ on $\alpha$ and $H_2:$ states that he is just a random individual in the population is:
\[
L(H_1,H_2|\alpha,\beta,\gamma) = \frac{P(D|\alpha,\beta,\gamma)}{P(D|\alpha,\gamma)}
\]
which can be simplified to:
\[
\lambda_j = \frac{P(\alpha_i,\beta_j,\gamma_k|\mathrm{paternity})}{P(\alpha_i,\beta_j,\gamma_k|\mathrm{non-paternity})} \\
= \frac{P(\beta_j)P(\alpha_i)T(\gamma_k|\alpha_i,\beta_j)}{P(\beta_j)P(\alpha_i)T(\gamma_k|\alpha_i)} \\
= \frac{T(\gamma_k|\alpha_i,\beta_j)}{T(\gamma_k|\alpha_i)}
\]
where $T(X|Y)$ is the Mendelian transition probability of offspring $X$ given parent $Y$.
Assumptions in Model of Paternity Likelihood
The basic paternity exclusion model assumes:
1. Completely random mating (can be modified by changin *priors*)
2. Independent assortment of alleles
Likelihood Example
Consider the maternal individual whose genotypes are:
\[
FP = \{AA,\;Bb,\;CC,\;Dd\}
\]
Whose $i^{th}$ offspring has the genotypes:
\[
O_i = \{AA,\;BB,\;Cc,\;dd\}
\]
Likelihood Example | $T(O|FP)$
The transition probability, $T(O|FP)$, is then:
Individual | Locus1 | Locus2 | Locus3 | Locus4
-----------|--------|--------|--------|--------
$FP$ | $AA$ | $Bb$ | $CC$ | $Dd$
$O_i$ | $AA$ | $BB$ | $Cc$ | $dd$
$T(O|FP) = 1*0.5*1*0.5 = 0.25$
Likelihood Example | Putative Male Parents
Individual | Locus1 | Locus2 | Locus3 | Locus4
-----------|--------|--------|--------|--------
$MP_1$ | $Aa$ | $BB$ | $cc$ | $Dd$
$MP_2$ | $AA$ | $BB$ | $Cc$ | $dd$
$\;$
Which one of the potential fathers is the <font class="alert">most likely</font> parent?
Likelihood Example | First Putative Father
Individual | Locus1 | Locus2 | Locus3 | Locus4
-------------|--------|--------|--------|--------
$FP$ | $AA$ | $Bb$ | $CC$ | $Dd$
$MP_1$ | $Aa$ | $BB$ | $cc$ | $Dd$
$O_i$ | $AA$ | $BB$ | $Cc$ | $dd$
$T(O|FP,MP)$ | 0.5 | 0.5 | 1.0 | 0.25
\[
T(O_1|FP,MP_1) = 0.5 * 0.5 * 1.0 * 0.25 = 0.0625
\]
And
\[
\lambda_1 = \frac{T(O_i|FP,MP_1)}{T(O_i|FP)} = \frac{0.0625}{0.25} = 0.25
\]
Likelihood Example | Second Putative Father
Individual | Locus1 | Locus2 | Locus3 | Locus4
-------------|--------|--------|--------|--------
$FP$ | $AA$ | $Bb$ | $CC$ | $Dd$
$MP_2$ | $AA$ | $BB$ | $Cc$ | $dd$
$O_i$ | $AA$ | $BB$ | $Cc$ | $dd$
$T(O|FP,MP)$ | 1.0 | 0.5 | 0.5 | 0.5
\[
T(O_1|FP,MP_2) = 0.5 * 0.5 * 1.0 * 0.25 = 0.125
\]
And
\[
\lambda_2 = \frac{T(O_i|FP,MP_1)}{T(O_i|FP)} = \frac{0.125}{0.25} = 0.5
\]
Likelihood Example | Interpretation of Results {.build}
Most likely parent is $MP_2$ because $\lambda_2 = 0.5 > \lambda_1 = 0.25$.
$\;$
Does this mean that $MP_2$ is the <font class="alert">real</font> parent?
Likelihood Example | In Class Exercise - Whose the daddies?
Individual | Locus 1 | Locus 2 | Locus 3
------------|----------|----------|---------
Mother | $A_1A_1$ | $B_1B_3$ | $C_1C_1$
Offspring 1 | $A_1A_2$ | $B_1B_3$ | $C_1C_2$
Offspring 2 | $A_1A_1$ | $B_3B_3$ | $C_1C_1$
Offspring 3 | $A_1A_1$ | $B_1B_1$ | $C_1C_1$
Dad 1 | $A_1A_2$ | $B_2B_3$ | $C_1C_1$
Dad 2 | $A_2A_2$ | $B_1B_1$ | $C_1C_2$
Dad 3 | $A_1A_1$ | $B_2B_3$ | $C_1C_2$
Dad 4 | $A_1A_1$ | $B_1B_1$ | $C_2C_2$
```{r}
library(gstudio)
loci <- c("Locus-A","Locus-B","Locus-C","Locus-D")
freqs <- data.frame(Locus = rep(loci, each = 4),
Allele = rep(LETTERS[1:4], times = 4),
Frequency = 0.25)
freqs
```
```{r}
adults <- make_population( freqs, N=100 )
adults$OffID <- 0
adults <- adults[ , c(1,6,2:5)]
adults[1:5,]
```
```{r}
offs <- data.frame()
mom <- adults[1,]
for( i in 1:20){
dad_id <- runif( 1, min=2, max=100)
dad <- adults[dad_id,]
off <- mate( mom, dad, N=1 )
offs <- rbind( offs, off )
}
offs$OffID <- 1:20
offs[1:5,]
```
```{r}
data <- rbind( adults, offs )
data <- data[ order(data$ID,data$OffID),]
rownames(data) <- 1:nrow(data)
data[1:10,]
```
```{r}
f <- frequencies( data[ data$OffID==0,] )
excl <- exclusion_probability( f )
excl
p <- excl$Pexcl
excl_multilocus <- 1 - prod( 1-p )
excl_multilocus
```
```{r}
family <- data[ data$ID==1, ]
minus_mom( family )
```
```{r}
dads <- adults[2:100,]
mom <- adults[1,]
off <- offs[1,]
for( i in 1:nrow(dads)){
dad <- dads[i,]
T <- transition_probability(off,mom,dad)
if( T > 0 )
cat("Father",i,"may be the real father (T =",T,")\n")
}
```
## Fractional Paternity
In cases where we have more than one putative father, we may want to get an idea of the relative strength of our inferences by comparing the likelihood ratios for all dads.
1. We may use arbitrary cut-offs, or
2. We may use all non-excluded dads, but weighted by their fractional contributions
Conditional Probability
**Problem:** We have several putative fathers ($MP_i, MP_j, MP_k, ... , MP_m$) have been found to have non-zero likelihoods of paternity.
$\;$
**Question:** What is the relative likelihood of paternity given these putative fathers?
Conditional Probability
Conditional probability determines the likelihood of an event (paternal likelihood) <font class="alert">given</font> that some other event has already happened (not excluded as a potential father).
\[
P(MP=j^*|FP=i,O=k) = \frac{P(O=k|FP=i,MP=j^*)P(MP=j^*|FP=i)}{\sum_{\forall j}P(O=k|FP=i,MP=j^*)P(MP=j^*|FP=i)}
\]
If we can assume that $P(MP=j|FP=i) = c$ (e.g., the frequencies of the female and male parents are constant with respect to the individual offspring being considered) then,
\[
P(MP=j^*|FP=i,O=k) = \frac{T(\gamma_k|\alpha_i,\beta_j^*)}{\sum_{\forall k}T(\gamma_k|\alpha_i,\beta_k^*)}
\]
Fractional Paternity
Some things to consider when using fractional analyses for paternity.
1. Not usually used in human studies.
2. Can be considered a *prior probability of paternity*.
3. Can include ecological, spatial, evolutionary components such as differential attractiveness, pollen fertility, output, etc.
4. Possible tautology
Every potential father is assigned paternity, the fraction of $X_{ik}$ on a particular $MP_j$ is proportional to the likelihood ratio.
```{r}
frac <- paternity( offs, mom, dads )
summary(frac)
frac[1:10,]
```
```{r}
t <- table(frac$OffID)
t
N_pexcl <- nrow(dads)*(1-excl_multilocus)
fit <- t.test(as.numeric(t),mu = N_pexcl)
fit
```
Maternity Analysis
Putative father identified by mother unknown.
\[
\lambda_i = \frac{P(\gamma|\beta)P(\beta)}{P(\gamma)P{\beta})} \\
= \frac{P(\gamma|\beta)}{P(\gamma)}
\]
where $P(\gamma)$ is the frequency of the offspring genotype in the population. All other things are the same.
Cryptic Gene Flow
Consider the case where:
1. You have identified a set of offspring collected from mothers.
2. Identified a set of fathers that are probabilistically sires of the offspring.
## Dispersal Kernels
Estimating the Disperal Distribution
Once a collection of paternity estimates have been determined, you can use them to estimate a *dispersal kernel*, describing the probability of paternity as a function of distance from the maternal individual.
```{r message=FALSE,echo=FALSE,fig.align='center'}
dist <- data.frame( Distance=rchisq(10000,df = 3))
ggplot(dist, aes(x=Distance)) + geom_density(fill="#99d8c9") + xlab("Physical Distance") + ylab("Probability of Paternity")
```
Dispersal Kernels | Distributions
The <font class="alert">form</font> of the distribution is critical for estimation. It determines:
> 1. The shape of the distribution
> 2. The variance of the distribution
> 3. Quantitative estimates and hypotheses you get from the data
Example Kernel Distribution Families
*Normal Family*
\[
p(a|x,y) = \frac{1}{\pi a^2}exp\left[-\left( \frac{r}{a} \right)^2\right]
\]
where $r = \sqrt{ x^2 + y^2}$ and $a = \sigma \sqrt{2}$.
$\;$
This produces a *thin tailed* distribution.
Example Kernel Distribution Families
*Exponential Family*
\[
p(a,b|x,y) = \frac{b}{2\pi a^2 \Gamma(2/b)}exp\left[ -\left( \frac{r}{a} \right)^b \right]
\]
where $\Gamma(a,b)$ is the gamma function and $b$ is a 'shape' parameter.
1. When $b=1$ This is the exponential distribution.
2. When $b=2$ this is the normal function.
3. When $b<1$ this is a fat-tailed distribution.
Example Kernel Distribution Families
Other distributions you may run across include:
1. The Geometric distribution,
2. The Weibull family of distributions,
3. The 2Dt family of distributions.
```{r}
r1 <- abs(rnorm(10000))
r2 <- rexp(10000)
df <- data.frame( Distribution=c(rep(c("Normal","Exponential"),each=10000)), Value=c(r1,r2))
library(ggplot2)
ggplot(df,aes(x=Value,fill=Distribution)) + geom_density(alpha=0.75) + theme_bw() + ylab("Frequency")
```
Concerns with kernel estimation
The following are some assumptions that are inherent in the use of dispersal kernels for estimating connectivity.
1. All functions are continuous,
2. All functions assume isotropy in dispersal,
3. All functions explicitly assume homogeneity of the dispersal matrix.
Skills
In this lecture we covered some rather simple parent/offspring relationships and how we can analyze them. Specifically, you should be comfortable with:
1. Understanding the qualities of loci that make for more powerful parentage analyses.
2. Be able to estimate single and multilocus exclusion probabilities and understand what they mean.
3. Estimate likelihood ratios for paternity given Mother, Offspring, and Putative Male Parent.
4. Use fractional paternity and understand conditional probability and how it applies to parentage.
5. Understand dispersal kernel estimation.