added boxplot for raw counts

graphics_bioinf.Rmd

@@ -389,7 +389,7 @@ dataL$locFit  <- predict(locfit(y~lp(a, nn=0.5, deg=1), data=dataL),
 
 #  Normalization and variance stabilization of bulk and single cell data
 
-## Size factors for bulk RNA--Seq
+## Size factors for  RNA--Seq data
 
 Before exploring normalization of single cell data, we look at this issue
 in bulk RNA--Seq: Next generation sequencing data is often analyzed in the
@@ -406,7 +406,7 @@ counts are comparable across samples. Such a factor is commonly called a
 __normalized sample counts__ = __raw sample counts__ / __ sample size factor__
 
 
-## A Colectoral Cancer exmaple data set
+## Normalization of a Colectoral Cancer exmaple data set
 
 We will illustrate this using a dataset from the [recount2 resource](http://dx.doi.org/10.1038/nbt.3838). [Löhr et. al., 2013](https://doi.org/10.1371/journal.pone.0067461) have 
 applied RNA-Seq  on colorectal normal,
@@ -460,17 +460,15 @@ Therefore, we can preview the actual count data matrix like so:
 ```{r pre_crc}
 counts_crc <- assay(crc_data)
 counts_crc[1:5, 1:5]
-
-counts_p4_meta <- counts_crc[, c("SRR837829", "SRR837847")]
-View(counts_p4_meta)
 ```
 
+
 The annotation for the genes looks like this:
 
 
 ```{r pre_crc_genes}
 colnames(colData(crc_data))
-colData(crc_data)[1:5, c("title", "characteristics", "mapped_read_count")]
+colData(crc_data)[1:5, c("title", "mapped_read_count")]
 nrow(colData(crc_data))
 ```
 
@@ -479,7 +477,42 @@ We have various normal, tumor and metastasis tissues from 8 donors. The authors
 had one pipelinr for the quantification of miRNA and one for the quantification 
 of "ordinary" genes. But as the recount2 ressource uses a unified pipeline with
 annotation data from the [genecode project](https://www.gencodegenes.org/stats/archive.html), which contains non coding features as well, we can focus on the samples annotated
-as "mRNA".
+as "mRNA". We will now create a column data table that contains the appropriate
+sample annotation.
+
+Specifically, we have to extract the sample annotations from 
+the title column and subset on the samples processed using mRNA 
+based quantification.
+
+```{r create_crc_col_data}
+
+col_data_crc  <- select(as.data.frame(colData(crc_data)),
+       title, characteristics, mapped_read_count) %>% 
+       rownames_to_column(var = "sample_id") %>%
+       as_tibble() %>%
+       tidyr::extract(title, into = c("quantification", "patient", "tissue"),
+               regex = "([[:alnum:]]+)_([[:alnum:]]+)_([[:alnum:]]+)") %>% 
+      dplyr::filter(quantification == "mRNA")
+      
+```
+
+We now plot the log counts of the mRNA samples to see whether their distributions
+are comparable. In order to avoid taking the log of zero, we only retain 
+a gene within a sample that has at least one count
+
+
+```{r mrna_counts}
+counts_crc_tidy <- rownames_to_column(data.frame(counts_crc), var = "ensembl_id") %>%
+                   as_tibble() %>%
+                   gather(key = "sample_id", value = "count", -ensembl_id) %>%
+                   dplyr::filter(sample_id %in% col_data_crc$sample_id) %>%
+                   dplyr::filter(count > 1)
+                   
+ggplot(counts_crc_tidy, aes(x = sample_id, y = log2(count), fill = sample_id) ) + 
+    geom_boxplot()  
+```
+
+
 
 <!-- the median gene expression per sample relative to a reference -->
 <!-- sample formed by the geometric mean of each single gene across samples. -->