Population Genomics Module 2

class: center, middle, inverse, title-slide

# Population Genomics Module 2
### Jinliang Yang
### Oct. 27, 2022

---

# Slides for today's class module

- Scan QR code to view the HTML slides:

- Lab website: https://jyanglab.com/
  - TEACHING: Courses
    - __Module 2__: [[HTML](https://jyanglab.com/slides/2022-module/week2/w2.html)]

---

# A recent paper on Nature Genetics

> [Chen et al., 2022](https://www.nature.com/articles/s41588-022-01184-y), Genome sequencing reveals evidence of adaptive variation in the genus _Zea_

- __744 genomes__, including maize and all wild taxa of the genus _Zea_

- Reveals __evidence of selection__ within taxa displaying __novel adaptation__

- Highlighted the __flowering-time-related pathways__ in the adaptation
  - Generate mutant alleles to validate two flowering-time candidate genes

---

# A recent paper on Nature Genetics

<div align="center">
<img src="ng2.png" height=200>
</div> 
> [Chen et al., 2022](https://www.nature.com/articles/s41588-022-01184-y), Genome sequencing reveals evidence of adaptive variation in the genus _Zea_

- #### Novel diversity in _Zea_
  - Heterozygosity and nucleotide diversity
  
- #### Signals of selection from allele frequency data
  - Using `\(F_{ST}\)`, XP-CLR

---

# Syllabus

### Module 1: Introduction and popgen terminology
- Introduction of population genomics
- Basic principles of evolutionary processes

### __Module 2: Diversity measurement__
- __Heterozygosity and diversity__
- __Population differentiation ( `\(F_{ST}\)` )__

### Module 3: Scan for direct and linked selection
- Direct selection
- Linked selection (XP-CLR)

---

# Polymorphism

Polymorphism is not always maintained by selection
> Kimura, 1962

### Mutation-drift equilibrium (in the absence of selection)

- Drift is reducing the number of alleles in the population

- While, mutation is increasing

- Eventually, equilibrium is reached with respect to the inbreeding coefficient ( `\(F\)` )

#### Inbreeding coefficient ( `\(F\)` )

Is the probability that the two alleles at any locus in an individual are identical by descent (IBD).
  - `\(F=0\)`, no inbreeding and `\(F=1\)`, completely inbreeding

---

# Mutation-drift equilibrium

- At equilibrium ( `\(F_e\)` ), `\(F_t = F_{t-1}\)`, with mutation rate of `\(\mu\)` and effective population size of `\(N_e\)`:

`\begin{align*}
F_e = \frac{(1 - \mu )^2}{2N_e - (2N_e -1)(1 - \mu)^2}
\end{align*}`

- But `\(\mu\)` is small as compared to 1 and can be ignored, so:

`\begin{align*}
F_e \approx \frac{1}{4N_e\mu + 1}
\end{align*}`

- `\(F_e\)` is identity (or homozygosity)

- So, the __expected heterozygosity__ at equilibrium is ( `\(1 - F_e\)` ) or

`\begin{align*}
1- F_e & \approx 1- \frac{1}{4N_e\mu + 1} \\
H_e  & \approx \frac{4N_e \mu}{4N_e\mu + 1}
\end{align*}`

---

# Mutation-drift equilibrium

`\begin{align*}
F_e \approx \frac{1}{4N_e\mu + 1}
\end{align*}`

- The relationship between `\(N_e\)`, `\(\mu\)`, and `\(F\)` at quilibrium.

```r
fe <- function(ne, mu){
  den <- 4*ne*mu + 1  # denominator
  return(1/den)
}
```

---

# Mutation-drift equilibrium

```r
ne <- seq(1, 50000, by=10)
plot(ne, fe(ne, mu=10^-4), type="l", lwd=3, col="red", 
     xlab="Effective population size (Ne)", ylab="F")
lines(ne, fe(ne, mu=10^-5), lty=2, lwd=3, col="blue")
lines(ne, fe(ne, mu=10^-8), lty=2, lwd=3, col="black")
```

---

# Mutation-drift equilibrium

.pull-left[

<div align="center">
<img src="ne_mu_f.png" width=450>
</div>
`\begin{align*}
1- F_e & \approx 1- \frac{1}{4N_e\mu + 1} \\
H_e  & \approx \frac{4N_e \mu}{4N_e\mu + 1}
\end{align*}`
]

.pull-right[
#### When the mutation rate is low (black line, `\(10^{-8}\)`), `\(F\)` is high even at very large Ne

- Mutations happen very infrequently and do not contribute much to the heterozygosity
  
  - Mutated alleles are easily lost through drift

#### When the mutation rate is relative high (__red line__, `\(10^{-4}\)`)

- Mutation counteracts drift through introducing new alleles at a high rate
]

---

# From Heterozygosity to Diversity

### Expected Heterozygosity ( `\(H_{exp}\)` )

- Probability that two randomly chosen copies of a locus are different alleles

- `\(H_{exp}\)` is often reported to __compare populations__

- Predict outlook for population or infer ancestral events
  
  - For example, high `\(H_{exp}\)` = maybe in HWE, large size, no obvious drift/inbreeding

`\(H_{exp}\)` = 1 - (avg expected __homozygosity__ over all loci)

`\begin{align*}
H_{exp} = 1 - \frac{1}{m}\sum_{k=l}^{m} \sum_{i=l}^{k} p_{ki}^2
\end{align*}`

- `\(m\)` is the number of loci
  - `\(k\)` is the number of alleles at a particular locus
  - `\(p_{ki}\)` is the frq of `\(i^{th}\)` allele at `\(k^{th}\)` locus

---

# Expected Heterozygosity

| Loci | Allele   | Freq   | 
| :-------: | : ------ : | :-------: | 
| locus A   | `\(A_1\)`    | `\(p_{11}=0.2\)`    | 
|           | `\(A_2\)`    | `\(p_{12}=0.8\)`    |
| locus B   |  `\(B_1\)`        |  `\(p_{21}=0.3\)`        |  
|           |  `\(B_2\)`        |  `\(p_{22}=0.7\)`        |

`\begin{align*}
H_{exp} = 1 - \frac{1}{m}\sum_{k=l}^{m} \sum_{i=l}^{k} p_{ki}^2
\end{align*}`

- `\(m=2\)` is the number of loci
  - `\(k=2\)` is the number of alleles at a particular locus
  - `\(p_{ki}\)` is the frq of `\(i^{th}\)` allele at `\(k^{th}\)` locus

`\begin{align*}
H_{exp} & = 1 - \frac{1}{2}\sum_{k=1}^{2} \sum_{i=1}^{2} p_{ki}^2 \\
& = 1 - \frac{1}{2}(0.2^2 + 0.8^2 + 0.3^2 + 0.7^2) \\
& = 0.37
\end{align*}`

---

# Expected Heterozygosity

- If `\(H_e\)` (or `\(H_{exp}\)`) is high, it does not mean that there are a lot of heterozygotes.

- It reflects potential diversity

- And how much __heterozygosity you would expect__ if the population were to randomly mate (no selection, migration, mutation, etc.).

# Diversity - allelic richness

- Count of how many alleles are in a sample

- Informative when comparing populations

- A population with fewer alleles may be prone to inbreeding
  
  - A locus with fewer alleles compared to others may indicate it is selected for/against (i.e., minor alleles tends to be deleterious alleles)

---

# Nucleotide Diversity ( `\(\pi\)` )

- Average of nucleotide difference between any two copies of a locus

- When you look at a nucleotide position of two random copies of a gene, diversity  ( `\(\pi\)` ) is the __probability they are different__

`\begin{align*}
\Pi = \frac{n}{n-1} \sum_{ij} x_i x_j \pi_{ij}
\end{align*}`

- `\(x_i\)` and `\(x_j\)` is the freq of the `\(i^{th}\)` and `\(j^{th}\)` haplotype
- `\(\pi_{ij}\)` is the proportion of nucleotide differences between `\(i^{th}\)` and `\(j^{th}\)` haplotype

---

# An example of Nucleotide Diversity

Here, we sequenced a population of 5 maize inbred lines for 500 bp and identified SNP variants as below.

--------------

- Total number of bp: `\(N_t = 500\)`
- Number of polymorphic: `\(N_p = 5\)`

- Proportion of polymorphic site: `\(P_n = \frac{N_p}{N_t} = \frac{5}{500} = 0.01\)`

---

# An example of Nucleotide Diversity

--------------

`\begin{align*}
\Pi = \frac{n}{n-1} \sum_{ij} x_i x_j \pi_{ij}
\end{align*}`

- `\(x_i\)` and `\(x_j\)` is the freq of the `\(i^{th}\)` and `\(j^{th}\)` haplotype
- `\(\pi_{ij}\)` is the proportion of nucleotide differences between `\(i^{th}\)` and `\(j^{th}\)` haplotype

---

# An example of Nucleotide Diversity

--------------

`\begin{align*}
\Pi = \frac{n}{n-1} \sum_{ij} x_i x_j \pi_{ij}
\end{align*}`

- `\(x_i\)` and `\(x_j\)` is the freq of the `\(i^{th}\)` and `\(j^{th}\)` haplotype
- `\(\pi_{ij}\)` is the proportion of nucleotide differences between `\(i^{th}\)` and `\(j^{th}\)` haplotype

- `\(\pi_{12} =3/500\)`, `\(\pi_{13} =3/500\)`, `\(\pi_{14} =3/500\)`, `\(\pi_{14} =2/500\)`
- `\(\pi_{23} =3/500\)`, `\(\pi_{24} =2/500\)`, `\(\pi_{14} =3/500\)`
- ...

---

# An example of Nucleotide Diversity

--------------

```r
df <- data.frame(hap=paste0("line", 1:5), snp1=c(0,0,1,1,1), 
                 snp2=c(1,1,0,1,1), snp3=c(0,1,1,1,0), 
                 snp4=c(1,0,1,0,1), snp5=c(0,1,1,0,1))
snp <- t(df[,-1])
getpi <- function(snp, totbp=500){
  nhap = ncol(snp) # number of hap
  td <- c()
  for(i in 1:(nhap -1)){
    for(j in (i+1):nhap){
      cn <- sum(abs(snp[,i] - snp[,j])) # count the nucleotide differences
      td <- c(td, cn)
    }
  }
  return(nhap/(nhap -1)*(1/nhap)^2*sum(td)/totbp)
}
```

---

# An example of Nucleotide Diversity

```r
getpi(snp, totbp=500)
```

```
## [1] 0.0028
```

Heterozygosity at this SNP site is:

`\begin{align*}
h = \frac{n}{n-1}(1- p^2 - q^2)
\end{align*}`

- `\(n\)` is the number of haplotypes in a population

Then the sum of site heterozygosities is:

`\begin{align*}
\sum_{i=1}^S h_i
\end{align*}`

- Here, `\(S\)` is the number of polymorphic sites
- `\(h_i\)` is the heterozygosity at the `\(i\)`th site

---

# An example of Nucleotide Diversity

--------------

```r
geth <- function(snp, totbp=500){
  nhap = ncol(snp) # number of hap
  q <- apply(snp, 1, sum)/nhap
  p <- 1 -q
  return(nhap/(nhap -1) * sum(rep(1, nhap) - p^2 - q^2)/totbp/2)
}
geth(snp, totbp=500)
```

```
## [1] 0.0028
```
---
# An example of Nucleotide Diversity

# Heterozygosity and Nucleotide Diversity

This estimate of heterozygosity gives us the average number of pairwise necleotide difference.

`\begin{align*}
\Pi = \sum_{i=1}^S h_i
\end{align*}`

---

# Population subdivision

.pull-left[
- ### Selection
- ### Migration
- ### Mutation
- ### Drift (inbreeding)
]

.pull-right[
<div align="center">
<img src="pops.png" width=400>
</div>
]

What if you did not know population boundaries before sampling?

### Cryptic population structure

Unknown (cryptic) population structure can appear as a __deficiency of heterozygotes__

---

# Population structure

### Wahlund Effect

- Unknown (cryptic) population structure can appear as __a deficiency of heterozygotes__

- Created reduced heterozygosity and increased homozygosity in your sample

---

# Population subdivision

| Genotype | `\(A_1A_1\)`   | `\(A_1A_2\)`    | `\(A_2A_2\)` 
| :-------: | : ------ : | :-------: | :-------: | 
| sub-pop1  | 64    | 32    | 4    | 
| sub-pop2  | 4     | 32    | 64    | 
| Total     | 68    | 64    | 68    |

The frequencies of the `\(A_1\)` allele in sub-pop1 and sub-pop2:
`\begin{align*}
p_1 & = \frac{2 \times 64 + 32}{2 \times 100} = 0.8 \\
p_2 & = \frac{2 \times 4 + 32}{2 \times 100} = 0.2 \\
\end{align*}`

The frequency of the `\(A_1\)` allele in the combined population:

`\begin{align*}
\bar{p} & = \frac{2 \times 68 + 64}{2 \times 200} = 0.5 \\
\end{align*}`

The expected value of heterozygotes for the combined population is:

`\begin{align*}
2\bar{p}\bar{q} \times n = 2 \times 0.5 \times 0.5 \times 200 = 100 \\ 
\end{align*}`

---

# Wahlund Effect

The _perceived deficiency_ of heterozygotes due to treating two different populations as one --- the __Wahlund effect__. 
> Wahlund, 1928

---

# Wahlund Effect

.pull-left[
<div align="center">
<img src="wahlund.png" height=400>
</div>
]

.pull-right[

- The blue dashed horizontal segment is the `\(H_{exp} = 2 \bar{p}\bar{q}\)`

- The red dashed horizontal segment is the `\(H_{obs}\)` 
  - mean value of the `\(H_1\)` and `\(H_2\)`

- Because the curve is concave downward, `\(H_{obs}\)` is always less than `\(H_{exp}\)`

]

Even if subpopulations are __partially isolated__, the Wahlund effect is held true.

---

# Heterozygosity to describe populations

In a subdivided population, the overall deviation from HWE heterozygosity:

- Factors acting within subpopulations
- Simply due to the subdivision itself (the Wahlund effect)

### Wright's F-statistics

Assuming HWE, Wright’s F describes the deviation from expected heterozygosity

`\begin{align*}
F = \frac{H_e - H_o}{H_e}
\end{align*}`

- Expected Heterozygosity ( `\(H_e\)` ) = estimate of diversity

- Observed Heterozygosity ( `\(H_o\)` )  = current characterization of the sample

- This measures the proportionate reduction in heterozygosity compared to HWE
  - If `\(H_o < H_e\)`, what does it mean and how can we use this?

---

# Heirarchical F-statistics

.pull-left[
<div align="center">
<img src="pops.png" width=300>
</div>
]

.pull-right[
<div align="center">
<img src="wahlund.png" height=280>
</div>
]

- `\(H_I\)` = mean observed heterozygosities over all subpopulations = `\(\frac{H_1 + H_2}{2}\)`

- `\(H_S\)` = mean expected heterozygosity within random mating subpopulations = `\(2p_iq_i\)`

- `\(H_T\)` = expected heterozygosity in a random mating total population = `\(2 \bar{p} \bar{q}\)`

---

# Hierarchical F-statistics

.pull-left[
<div align="center">
<img src="wahlund.png" height=280>
</div>
]

.pull-right[
- `\(H_I\)`, the average of the observed heterozygosities

- `\(H_S\)`, the average of the expected heterozygosities
]

The average deviation in heterozygosity within subpopulations is

`\begin{align*}
F_{IS} = \frac{H_S - H_I}{H_S}
\end{align*}`

- The subscript `\(IS\)` here indicates deviation among __individuals__ relative to their __subpopulation__

---

# Hierarchical F-statistics

.pull-left[
<div align="center">
<img src="wahlund.png" height=280>
</div>
]

.pull-right[
- `\(H_I\)`, the average of the observed heterozygosities

- `\(H_S\)`, the average of the expected heterozygosities

- `\(H_T\)`, the expected heterozygosity in the toal population
]

The deviation in heterozygosity due to subdivision alone is

`\begin{align*}
F_{ST} = \frac{H_T - H_S}{H_T}
\end{align*}`

- The subscript `\(ST\)` here indicates deviation among __subpopulations__ relative to the __total population__

---

# Hierarchical F-statistics

.pull-left[
<div align="center">
<img src="wahlund.png" height=280>
</div>
]

.pull-right[
- `\(H_I\)`, the average of the observed heterozygosities

- `\(H_S\)`, the average of the expected heterozygosities

- `\(H_T\)`, the expected heterozygosity in the toal population
]

Finally, the overall deviation in heterozygosity in the total pouplation is

`\begin{align*}
F_{IT} = \frac{H_T - H_I}{H_T}
\end{align*}`

- The subscript `\(IT\)` here indicates deviation among __individuals__ relative to the __total population__

---

# Summary of Hierarchical F-statistics

### `\(F_{ST}\)`

- Measures how different the subpopulations are

- It can be interpreted as the proportion of the total heterozygosity that is due to differences in allele frequencies among subpopulations

- `\(F_{ST}\)` can be calculated from allele frequencies alone

### `\(F_{IS}\)` and `\(F_{IT}\)`

- `\(F_{IS}\)` is due to factors (selection, drift, mutation) acting within subpopulations

- `\(F_{IT}\)` depends on the values of `\(F_{ST}\)` and `\(F_{IS}\)`

- `\(F_{IS}\)` and `\(F_{IT}\)` require observed genotype frequencies

---

# An example for F-statistics calculation

Heterozygosity indicies (over individuals `\(H_{I}\)`, subpopulations `\(H_{S}\)`, and total population `\(H_{T}\)`)

- `\(H_I\)` based on __observed__ heterozygosities in __subpopulations__:

`\begin{align*}
H_I & = (H_{osb1} \times N_1 + H_{obs2} \times N_2) \times \frac{1}{N_{total}} \\
& = (0.32 \times 100 + 0.32 \times 100) \times \frac{1}{200} = 0.32 \\
\end{align*}`

---

# An example for F-statistics calculation

The frequencies of the `\(A_1\)` allele in sub-pop1, sub-pop2, and combined population:
`\begin{align*}
p_1  = 0.8, p_2 = 0.2, \bar{p} = 0.5 \\
\end{align*}`

- `\(H_S\)` based on __expected__ heterozygosities in __subpopulations__:

`\begin{align*}
H_S & = (H_{exp1} \times N_1 + H_{exp2} \times N_2) \times \frac{1}{N_{total}} \\
& = (2 \times 0.8 \times 0.2 \times 100 + 2 \times 0.8 \times 0.2 \times 100) \times \frac{1}{200} = 0.32 \\
\end{align*}`

- `\(H_T\)` based on __expected__ heterozygosities in __total populations__:

`\begin{align*}
H_T & = 2 \bar{p} \bar{q} \\
& = 2 \times 0.5 \times 0.5 = 0.5 \\
\end{align*}`

---

# An example for F-statistics calculation

`\begin{align*}
H_I & = 0.32 \\
H_S & = 0.32 \\
H_T & = 0.5 \\
\end{align*}`

--------------

`\begin{align*}
F_{IS} & = \frac{H_S - H_I}{H_S}  = \frac{0.32 - 0.32}{0.32} =0 \\
\end{align*}`

- There is no evidence of inbreeding within each subpopulations

`\begin{align*}
F_{ST} & = \frac{H_T - H_S}{H_T}  = \frac{0.5 - 0.32}{0.5} = 0.36 \\
\end{align*}`

`\begin{align*}
F_{IT} & = \frac{H_T - H_I}{H_T}  = \frac{0.5 - 0.32}{0.5} = 0.36 \\
\end{align*}`

- But clear evidence of population subdivision

- And population subdivision accounts for all the genetic variation

---
# More on the Hierarchical F-statistics

### `\(F_{IS}\)` - Inbreeding index

- Reduction in heterozygosity of an individual due to non-random mating in a subpopulation

### `\(F_{ST}\)`  - Fixation index

- Reduction in heterozygosity of a subpopulation relative to the total population due to drift;

- Measure of genetic differentiation

---
# More on the Hierarchical F-statistics

### `\(F_{IT}\)` - Overall fixation index

- Mean reduction in heterozygosity of an individual relative to the total population

- Combined effects of non-random mating ( `\(F_{IS}\)` ) with drift ( `\(F_{ST}\)` )

### Relationship of `\(F_{IS}\)`, `\(F_{ST}\)`, and `\(F_{IT}\)` 
However, it is NOT simple additive relationship

They are related by the expression
`\begin{align*}
(1-F_{IT}) = (1-F_{IS})(1-F_{ST})
\end{align*}`

It can be rewritten as

`\begin{align*}
F_{IT} = F_{IS} + F_{ST} - F_{IS}F_{ST}
\end{align*}`

---

# Of course there are issues...

`\(F_{ST}\)` must be adjusted to account for the numbers of individuals in each sub-population (sampling error)

- Unbiased estimators, such as `\(\hat{F_{ST}}\)`
  > Nei and Chesser, 1983

CANNOT compare `\(F_{ST}\)` between studies or samples if they are not based upon the exact same loci!!

- You can express `\(F_{ST}\)` as `\(F_{ST}'\)` --- a normalized version
> Hedrick, 2005

- Jost’s D uses number of effective alleles instead of heterozygosity
> Jost, 2008