Lesson 4 Exercise Questions: Tidyverse

The filtlowabund_scaledcounts_airways.txt includes normalized and non-normalized transcript count data from an RNAseq experiment. You can read more about the experiment here. You can obtain the data outside of class here.

The diffexp_results_edger_airways.txt includes results from differential expression analysis using EdgeR. You can obtain the data outside of class here.

Putting what we have learned to the test:
The following questions synthesize several of the skills you have learned thus far. It may not be immediately apparent how you would go about answering these questions. Remember, the R community is expansive, and there are a number of ways to get help including but not limited to google search. These questions have multiple solutions, but try to solve the problem using tidyverse.

The normalized and non-normalized count data should be saved to the object scaled_counts. The differential expression results should be saved to the object dexp.

Select the following columns from the scaled_counts data frame: sample, cell, dex, Run, transcript, avgLength, and counts_scaled. However, rearrange the columns so that the column 'Run' follows 'sample' and 'avgLength' is the last column. Save this to the object df_counts.
Solution
```
df_counts<-scaled_counts %>%
  select(sample, Run, cell, dex, transcript,counts_scaled,avgLength) 
```

Explore the column 'avgLength' in df_counts. Does the data in this column vary within a sample? How could we figure this out if we didn't know what was in this column?

Solution

df_counts %>% group_by(sample) %>% summarize(median=median(avgLength),
                                            max=max(avgLength),
                                            min=min(avgLength))

Using the differential expression results, create a data frame with the top five differentially expressed genes by p-value. Hint: Top genes in this case will have the smallest FDR corrected p-value and an absolute value of the log fold change greater than 2. (Lesson 4 challenge question)
Solution
```
topgene<-dexp %>%
  arrange(FDR) %>%
  filter(logFC >= abs(2)) %>%
  head(5)
```
Filter the data frame scaled_counts to include only our top five differentially expressed genes (from question 3) and save to a new object named top_gene_counts.
Solution
```
top_gene_counts<-
  scaled_counts %>%
  filter(transcript %in% topgene$transcript) 
```

Create a data frame of the mean, median, and standard deviation of the normalized counts for each of our top transcripts by treatment (dex). Is there a large amount of variation within a treatment?

Solution

top_gene_counts %>%
  group_by(dex, transcript) %>% 
  summarize(mean_counts=mean(counts_scaled),
            sd=sd(counts_scaled),
            median=median(counts_scaled))

Return a filtered data frame of the differential expression results. We want to look at only the transcripts with logCPM greater than 3 with a logFC greater than or equal to an absolute value of 2.5 and an adjusted (FDR) p-value less than 0.001.
Solution
```
dexp %>%
  filter(logCPM > 3,logFC >= abs(2.5), FDR < 0.001) 
```