You only need to write lines of code for each question. When answering questions that ask you to identify or interpret something, the length of your response doesn’t matter. For example, if the answer is just ‘yes,’ ‘no,’ or a number, you can just give that answer without adding anything else.
+
We will go through comparable code and concepts in the live learning session. If you run into trouble, start by using the help help() function in R, to get information about the datasets and function in question. The internet is also a great resource when coding (though note that no outside searches are required by the assignment!). If you do incorporate code from the internet, please cite the source within your code (providing a URL is sufficient).
+
Please bring questions that you cannot work out on your own to office hours, work periods or share with your peers on Slack. We will work with you through the issue.
+
You will need to install PLINK and run the analyses. Please follow the OS-specific setup guide in SETUP.md. PLINK is a free, open-source whole genome association analysis toolset, designed to perform a range of basic, large-scale analyses in a computationally efficient manner.
+
+
Question 1: Data inspection
+
Before fitting any models, it is essential to understand the data. Use R or bash code to answer the following questions about the gwa.qc.A1.fam, gwa.qc.A1.bim, and gwa.qc.A1.bed files, available at the following Google Drive link: https://drive.google.com/drive/folders/11meVqGCY5yAyI1fh-fAlMEXQt0VmRGuz?usp=drive_link. Please download all three files from this link and place them in 02_activities/data/.
+
+
# Load the packages needed for this assignment
+library(data.table)
+library(ggplot2)
+library(seqminer)
+library(HardyWeinberg)
+library(dplyr)
+
+
+
Read the .fam file. How many samples does the dataset contain?
+
+
+
wc-l ../data/gwa.qc.A1.fam
+
+
4000 ../data/gwa.qc.A1.fam
+
+
+
The fam dataset contains 4000 samples
+
+
What is the ‘variable type’ of the response variable (i.e.Continuous or binary)?
Read the .bim file. How many SNPs does the dataset contain?
+
+
+
wc-l ../data/gwa.qc.A1.bim
+
+
101083 ../data/gwa.qc.A1.bim
+
+
+
The bim file has 101083 SNPs
+
+
+
Question 2: Allele Frequency Estimation
+
+
Load the genotype matrix for SNPs rs1861, rs3813199, rs3128342, and rs11804831 using additive coding. What are the allele frequencies (AFs) for these four SNPs?
+
+
+
+# Create SNP list
+printf"%s\n" rs1861 rs3813199 rs3128342 rs11804831 > ../data/snplist_A1.txt
+cat ../data/snplist_A1.txt
+
+# Subset the 4 SNPs from the PLINK dataset
+plink2--bfile ../data/gwa.qc.A1 --extract ../data/snplist_A1.txt --make-bed--out ../data/gwa_A1_subset
+
+# Additive coding on the subsetted SNPs
+plink2--bfile ../data/gwa_A1_subset --export A --out ../data/gwa_A1_subset_additive
+
+# Calculate allele frequencies for the 4-SNP subset
+plink2--bfile ../data/gwa_A1_subset --freq--out ../data/gwa_A1_subset_freq
Since the estimated allele frequencies of these SNPs are <0.5 from the ALT_FREQS column of the PLINK frequency output, ALT is already the minor allele. Therefore, ALT_FREQS = minor allele frequencies.
+
+
+
Question 3: Hardy–Weinberg Equilibrium (HWE) Test
+
+
Conduct the Hardy–Weinberg Equilibrium (HWE) test for all SNPs in the .bim file. Then, load the file containing the HWE p-value results and display the first few rows of the resulting data frame.
The HWE p-values for the four SNPs are as follows: rs1861 = 0.274719, rs3813199 = 1.000000, rs3128342 = 0.330273, rs11804831 = 0.113354
+
+
+
Question 4: Genetic Association Test
+
+
Conduct a linear regression to test the association between SNP rs1861 and the phenotype. What is the p-value?
+
+
+
# Linear regression model
+geno <-fread("../data/gwa_A1_subset_additive.raw")
+model <-lm(PHENOTYPE ~ rs1861_C, data = geno)
+summary(model)
+
+
+Call:
+lm(formula = PHENOTYPE ~ rs1861_C, data = geno)
+
+Residuals:
+ Min 1Q Median 3Q Max
+-3.5439 -0.6850 0.0021 0.6993 3.3268
+
+Coefficients:
+ Estimate Std. Error t value Pr(>|t|)
+(Intercept) 0.05238 0.09486 0.552 0.581
+rs1861_C 0.97382 0.04943 19.703 <2e-16 ***
+---
+Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
+
+Residual standard error: 1.003 on 3962 degrees of freedom
+ (36 observations deleted due to missingness)
+Multiple R-squared: 0.08924, Adjusted R-squared: 0.08901
+F-statistic: 388.2 on 1 and 3962 DF, p-value: < 2.2e-16
+
+
+
The p-value for the association between SNP rs1861 and the phenotype is <2e-16.
+
+
How would you interpret the beta coefficient from this regression?
+
+
The regression coefficient for rs1861 is 0.97382 (p <2e-16), indicating that each additional copy of the C allele is associated with an increase of approximately 0.97 units of the phenotype on average. The significant p-value suggests evidence of an association between this SNP and the phenotype in our sample.
+
+
Plot the scatterplot of phenotype versus the genotype of SNP rs1861. Add the regression line to the plot.
+
+
+
# Load the genotype data
+geno <-fread("../data/gwa_A1_subset_additive.raw")
+
+# Remove missing values before plotting
+plot_data_add <- geno[!is.na(rs1861_C) &!is.na(PHENOTYPE), ]
+
+# Create scatterplot with regression line assuming outcome is phenotype and predictor is genotype of SNP rs1861
+p <-ggplot(plot_data_add, aes(x =factor(rs1861_C), y = PHENOTYPE)) +
+geom_point(color ="blue", alpha =0.5) +
+geom_smooth(aes(group =1), method ="lm", color ="red") +
+labs(
+title ="Scatterplot showing the association between SNP rs1861 and phenotype",
+x ="Genotype (number of C alleles)",
+y ="Phenotype"
+ ) +
+theme_minimal()
+
+p
+
+
+
+
+
+
+
+
+
+
Convert the genotype coding for rs1861 to recessive coding.
Conduct a linear regression to test the association between the recessive-coded rs1861 and the phenotype. What is the p-value?
+
+
+
# Linear regression model using recessive coding
+model_recessive <-lm(PHENOTYPE ~ rs1861_recessive, data = geno_recessive)
+summary(model_recessive)
+
+
+Call:
+lm(formula = PHENOTYPE ~ rs1861_recessive, data = geno_recessive)
+
+Residuals:
+ Min 1Q Median 3Q Max
+-3.5437 -0.6892 0.0015 0.7016 3.3270
+
+Coefficients:
+ Estimate Std. Error t value Pr(>|t|)
+(Intercept) 0.99231 0.04945 20.07 <2e-16 ***
+rs1861_recessive 1.00754 0.05224 19.29 <2e-16 ***
+---
+Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
+
+Residual standard error: 1.005 on 3962 degrees of freedom
+ (36 observations deleted due to missingness)
+Multiple R-squared: 0.08582, Adjusted R-squared: 0.08559
+F-statistic: 372 on 1 and 3962 DF, p-value: < 2.2e-16
+
+
+
The p-value for the association between SNP rs1861 and the phenotype using recessive coding is <2e-16.
+
+
Plot the scatterplot of phenotype versus the recessive-coded genotype of rs1861. Add the regression line to the plot.
+
+
+
# Create scatterplot with regression line assuming outcome is phenotype and predictor is genotype of SNP rs1861 using recessive coding
+plot_data_rec <- geno_recessive[!is.na(rs1861_recessive) &!is.na(PHENOTYPE), ]
+
+p_rec <-ggplot(plot_data_rec,
+aes(x =factor(rs1861_recessive), y = PHENOTYPE)) +
+geom_point(color ="blue", alpha =0.5) +
+geom_smooth(aes(group =1), method ="lm", color ="red") +
+labs(
+title ="Scatterplot showing the association between recessive-coded rs1861 \nand phenotype",
+x ="Recessive genotype (1 = two copies of allele)",
+y ="Phenotype"
+ ) +
+theme_minimal()
+
+p_rec
+
+
+
+
+
+
+
+
+
+
Which model fits better? Justify your answer.
+
+
+
AIC(model)
+
+
[1] 11276.91
+
+
AIC(model_recessive)
+
+
[1] 11291.74
+
+
+
The additive model fits better than the recessive model. In the linear regression analyses, the additive model had a higher R-squared value compared to the recessive model (0.08924 vs. 0.08582). Furthermore, the AIC of the additive model (11276.91) is lower than the AIC of the recessive model (11291.74), indicating that the additive coding provides a better overall model fit. This suggests that the effect of rs1861 on the phenotype is better captured by an allele-dose (additive) effect rather than a recessive effect.
+
+
+
Criteria
+
+
+
+
+
+
+
+
+
Criteria
+
Complete
+
Incomplete
+
+
+
+
+
Data Inspection
+
Correct sample/SNP counts and variable type identified.
+
Missing or incorrect counts or variable type.
+
+
+
Allele Frequency Estimation
+
Correct allele and minor allele frequencies computed.
+
Frequencies missing or wrong.
+
+
+
Hardy–Weinberg Equilibrium Test
+
Correct PLINK command and p-value extraction in R.
+
PLINK command or extraction incorrect/missing.
+
+
+
Genetic Association Test
+
Correct regressions, plots, coding, and interpretation.
+
Regression, plots, or interpretation missing/incomplete.
+
+
+
+
+
+
Submission Information
+
📌 Please review our Assignment Submission Guide for detailed instructions on how to format, branch, and submit your work. Following these guidelines is crucial for your submissions to be evaluated correctly.
+
+
Note:
+
If you like, you may collaborate with others in the cohort. If you choose to do so, please indicate with whom you have worked with in your pull request by tagging their GitHub username. Separate submissions are required.
+
+
+
+
Submission Parameters
+
+
Submission Due Date: 11:59 PM – 16/03/2026
+
Branch name for your repo should be: assignment-1
+
What to submit for this assignment:
+
+
Populate this Quarto document (assignment_1.qmd).
+
Render the document with Quarto: quarto render assignment_1.qmd.
+
Submit both assignment_1.qmd and the rendered HTML file assignment_1.html in your pull request.
+
+
What the pull request link should look like for this assignment: https://github.com/<your_github_username>/gen_data/pull/<pr_id>
+
+
Open a private window in your browser. Copy and paste the link to your pull request into the address bar. Make sure you can see your pull request properly. This helps the technical facilitator and learning support team review your submission easily.
+
+
+
+
Checklist:
+
+
Created a branch with the correct naming convention.
+
Ensured that the repository is public.
+
Reviewed the PR description guidelines and adhered to them.
+
Verified that the link is accessible in a private browser window.
+
Confirmed that both assignment_1.qmd and assignment_1.html are included in the pull request.
+
+
If you encounter any difficulties or have questions, please don’t hesitate to reach out to our team via our Slack help channel. Our technical facilitators and learning support team are here to help you navigate any challenges.
+
+
+
+
+
+
+
+
+
+
+
+
+
\ No newline at end of file
diff --git a/02_activities/assignments/assignment_1.qmd b/02_activities/assignments/assignment_1.qmd
index 550af3d..f0bdd0b 100644
--- a/02_activities/assignments/assignment_1.qmd
+++ b/02_activities/assignments/assignment_1.qmd
@@ -17,96 +17,200 @@ You will need to install PLINK and run the analyses. Please follow the OS-specif
Before fitting any models, it is essential to understand the data. Use R or bash code to answer the following questions about the `gwa.qc.A1.fam`, `gwa.qc.A1.bim`, and `gwa.qc.A1.bed` files, available at the following Google Drive link: . Please download all three files from this link and place them in `02_activities/data/`.
-(i) Read the .fam file. How many samples does the dataset contain?
+```{r, message=FALSE, warning=FALSE, results='hide'}
+# Load the packages needed for this assignment
+library(data.table)
+library(ggplot2)
+library(seqminer)
+library(HardyWeinberg)
+library(dplyr)
+```
-```
-# Your answer here...
+(i) Read the .fam file. How many samples does the dataset contain?
+
+```{bash}
+wc -l ../data/gwa.qc.A1.fam
```
+The fam dataset contains 4000 samples
+
(ii) What is the 'variable type' of the response variable (i.e.Continuous or binary)?
-```
-# Your answer here...
+```{bash}
+head ../data/gwa.qc.A1.fam
```
+The variable type or phenotype is continuous
+
(iii) Read the .bim file. How many SNPs does the dataset contain?
-```
-# Your answer here...
+```{bash}
+wc -l ../data/gwa.qc.A1.bim
```
+The bim file has 101083 SNPs
+
#### Question 2: Allele Frequency Estimation
(i) Load the genotype matrix for SNPs rs1861, rs3813199, rs3128342, and rs11804831 using additive coding. What are the allele frequencies (AFs) for these four SNPs?
-```
-# Your code here...
+```{bash,results='hide',message=FALSE, warning=FALSE}
+
+# Create SNP list
+printf "%s\n" rs1861 rs3813199 rs3128342 rs11804831 > ../data/snplist_A1.txt
+cat ../data/snplist_A1.txt
+
+# Subset the 4 SNPs from the PLINK dataset
+plink2 --bfile ../data/gwa.qc.A1 --extract ../data/snplist_A1.txt --make-bed --out ../data/gwa_A1_subset
+
+# Additive coding on the subsetted SNPs
+plink2 --bfile ../data/gwa_A1_subset --export A --out ../data/gwa_A1_subset_additive
+
+# Calculate allele frequencies for the 4-SNP subset
+plink2 --bfile ../data/gwa_A1_subset --freq --out ../data/gwa_A1_subset_freq
```
+```{r,message=FALSE, warning=FALSE}
+# Load additive-coded genotype matrix
+geno <- fread("../data/gwa_A1_subset_additive.raw")
+head(geno)
+```
+
+```{r,message=FALSE, warning=FALSE}
+# Read and display allele frequencies of four SNPs
+freq <- fread("../data/gwa_A1_subset_freq.afreq")
+af_table <- freq[, .(SNP = ID, AF = ALT_FREQS)]
+af_table
+```
+
+The allele frequencies (AF) of the four SNPs are as follows: rs1861 = 0.0539859, rs3813199 = 0.0569126, rs3128342 = 0.3051210, rs11804831 = 0.1543410
+
(ii) What are the minor allele frequencies (MAFs) for these four SNPs?
-```
-# Your code here...
+```{r,message=FALSE, warning=FALSE}
+maf_table <- freq[, .(SNP = ID, AF = ALT_FREQS, MAF = pmin(ALT_FREQS, 1 - ALT_FREQS))]
+maf_table
```
+Since the estimated allele frequencies of these SNPs are <0.5 from the ALT_FREQS column of the PLINK frequency output, ALT is already the minor allele. Therefore, ALT_FREQS = minor allele frequencies.
+
#### Question 3: Hardy–Weinberg Equilibrium (HWE) Test
(i) Conduct the Hardy–Weinberg Equilibrium (HWE) test for all SNPs in the .bim file. Then, load the file containing the HWE p-value results and display the first few rows of the resulting data frame.
-```
-# Your code here...
+```{bash,results='hide',message=FALSE, warning=FALSE}
+plink2 --bfile ../data/gwa.qc.A1 --hardy --out ../data/gwa_qc_A1_hwe
+```
+
+```{r, message=FALSE, warning=FALSE}
+hwe <- fread("../data/gwa_qc_A1_hwe.hardy")
+head(hwe)
```
(ii) What are the HWE p-values for SNPs rs1861, rs3813199, rs3128342, and rs11804831?
-```
-# Your code here...
+```{r, message=FALSE, warning=FALSE}
+# Create a subset of the four SNPs
+snps_interest <- c("rs1861", "rs3813199", "rs3128342", "rs11804831")
+hwe_subset <- hwe[ID %in% snps_interest, .(SNP = ID, HWE_P = P)]
+hwe_subset
```
+The HWE p-values for the four SNPs are as follows: rs1861 = 0.274719, rs3813199 = 1.000000, rs3128342 = 0.330273, rs11804831 = 0.113354
+
#### Question 4: Genetic Association Test
(i) Conduct a linear regression to test the association between SNP rs1861 and the phenotype. What is the p-value?
-```
-# Your code here...
+```{r, message=FALSE, warning=FALSE}
+# Linear regression model
+geno <- fread("../data/gwa_A1_subset_additive.raw")
+model <- lm(PHENOTYPE ~ rs1861_C, data = geno)
+summary(model)
```
+The p-value for the association between SNP rs1861 and the phenotype is <2e-16.
+
(ii) How would you interpret the beta coefficient from this regression?
-```
-# Your answer here...
-```
+The regression coefficient for rs1861 is 0.97382 (p <2e-16), indicating that each additional copy of the C allele is associated with an increase of approximately 0.97 units of the phenotype on average. The significant p-value suggests evidence of an association between this SNP and the phenotype in our sample.
(iii) Plot the scatterplot of phenotype versus the genotype of SNP rs1861. Add the regression line to the plot.
-```
-# Your code here...
+```{r, message=FALSE, warning=FALSE}
+# Load the genotype data
+geno <- fread("../data/gwa_A1_subset_additive.raw")
+
+# Remove missing values before plotting
+plot_data_add <- geno[!is.na(rs1861_C) & !is.na(PHENOTYPE), ]
+
+# Create scatterplot with regression line assuming outcome is phenotype and predictor is genotype of SNP rs1861
+p <- ggplot(plot_data_add, aes(x = factor(rs1861_C), y = PHENOTYPE)) +
+ geom_point(color = "blue", alpha = 0.5) +
+ geom_smooth(aes(group = 1), method = "lm", color = "red") +
+ labs(
+ title = "Scatterplot showing the association between SNP rs1861 and phenotype",
+ x = "Genotype (number of C alleles)",
+ y = "Phenotype"
+ ) +
+ theme_minimal()
+
+p
```
(iv) Convert the genotype coding for rs1861 to recessive coding.
-```
-# Your code here...
+```{r, message=FALSE, warning=FALSE}
+# Load genotype data
+geno <- fread("../data/gwa_A1_subset_additive.raw")
+
+# Convert coding for rs1861 to recessive coding
+geno_recessive <- geno
+geno_recessive$rs1861_recessive <- ifelse(geno_recessive$rs1861_C == 2, 1, 0)
+
+# Check recoding
+table(geno_recessive$rs1861_C, geno_recessive$rs1861_recessive, useNA = "ifany")
```
(v) Conduct a linear regression to test the association between the recessive-coded rs1861 and the phenotype. What is the p-value?
-```
-# Your code here...
+```{r, message=FALSE, warning=FALSE}
+# Linear regression model using recessive coding
+model_recessive <- lm(PHENOTYPE ~ rs1861_recessive, data = geno_recessive)
+summary(model_recessive)
```
+The p-value for the association between SNP rs1861 and the phenotype using recessive coding is <2e-16.
+
(vi) Plot the scatterplot of phenotype versus the recessive-coded genotype of rs1861. Add the regression line to the plot.
-```
-# Your code here...
+```{r, message=FALSE, warning=FALSE}
+# Create scatterplot with regression line assuming outcome is phenotype and predictor is genotype of SNP rs1861 using recessive coding
+plot_data_rec <- geno_recessive[!is.na(rs1861_recessive) & !is.na(PHENOTYPE), ]
+
+p_rec <- ggplot(plot_data_rec,
+ aes(x = factor(rs1861_recessive), y = PHENOTYPE)) +
+ geom_point(color = "blue", alpha = 0.5) +
+ geom_smooth(aes(group = 1), method = "lm", color = "red") +
+ labs(
+ title = "Scatterplot showing the association between recessive-coded rs1861 \nand phenotype",
+ x = "Recessive genotype (1 = two copies of allele)",
+ y = "Phenotype"
+ ) +
+ theme_minimal()
+
+p_rec
```
(vii) Which model fits better? Justify your answer.
-```
-# Your answer here...
+```{r, message=FALSE, warning=FALSE}
+AIC(model)
+AIC(model_recessive)
```
+The additive model fits better than the recessive model. In the linear regression analyses, the additive model had a higher R-squared value compared to the recessive model (0.08924 vs. 0.08582). Furthermore, the AIC of the additive model (11276.91) is lower than the AIC of the recessive model (11291.74), indicating that the additive coding provides a better overall model fit. This suggests that the effect of rs1861 on the phenotype is better captured by an allele-dose (additive) effect rather than a recessive effect.
+
### Criteria
| Criteria | Complete | Incomplete |
