Differential Gene Expression

Two outputs from Cuffdiff are used in this analyse:

## [1] "gene_exp.diff"       "genes.fpkm_tracking"

1. Firstly I drawn a scatter plot of FPKM to check gene’s expression among conditions.

2. Secondly a density plot was drawn to compare the whole genes’ expression trend. x-axis is the gene expression in log10(FPKM); y-axis is the density for every expression level. Generally M1 and M2 marcophages’ expression patterns are more similiar to each other, compared with Monocytes. For the gene expression, there are two peaks, around Log10(FPKM) -1 and 1 separately. Upon M1/M2 stimulation, there are more genes in the higher peak, and the lower peak shifted to more lower part.

To further check whether “the genes shifted to more lower expression” come from the lower peak or higher peak, I split the Monocyte genes into two population, log10(FPKM)>=0 and log10(FPKM)<0. And draw density plot separetely. Note: Because I show the group trend here, there is no sharp line on log10(FPKM) 0, but a slope.

As we can see, For the genes in higher peak of Monocyte (upper figure,log10(FPKM)>=0), there are more down-regulated genes than up-regulated. For genes lower peak of Monocyte, (lower figure, log10(FPKM)<0), some genes’ expression is shifted to more lower part, and some are up-regulated to the higher peak.

3. Then I draw a volcano plot to compare the significant DEGs, with non-significant genes.

Differential Alternative Splicing Events (DASE)

Only rMATS outputs are used for the analysis, because when I run the software, rMATS is the only one which can handle replicates. There are two sets of output from rMATS: Junction-Count-Only and Reads-On-Target-And-Junction-Counts, depending on whether use the reads only mapped to the target exons or not. Take the example of Monocytes/M1-macrophage, the output lists are:

##  [1] "A3SS.MATS.JunctionCountOnly.txt"             
##  [2] "A3SS.MATS.ReadsOnTargetAndJunctionCounts.txt"
##  [3] "A5SS.MATS.JunctionCountOnly.txt"             
##  [4] "A5SS.MATS.ReadsOnTargetAndJunctionCounts.txt"
##  [5] "MXE.MATS.JunctionCountOnly.txt"              
##  [6] "MXE.MATS.ReadsOnTargetAndJunctionCounts.txt" 
##  [7] "RI.MATS.JunctionCountOnly.txt"               
##  [8] "RI.MATS.ReadsOnTargetAndJunctionCounts.txt"  
##  [9] "SE.MATS.JunctionCountOnly.txt"               
## [10] "SE.MATS.ReadsOnTargetAndJunctionCounts.txt"

1. Firstly, I compared the data of these two sets, to see which is better.

There are no clear differences between these two calculating methods. Even the significant event numbers are similar, Junction-only 4508, and Junction-Target 4345. So I use Reads-On-Target-And-Junction-Counts in the following analysis.

2. Second, similar to the DGE, I plot PSI density of all the events and DASEs separately. The lower plot is zoomed in of the upper plot (PSI 0-0.96 part). There are a large amout of exons, whose inclusion ratio is close to 1, means they are constitutive. In the significant events, these genes are much lower. There are also a big population of exons, whose inclusion ratio around 0.5, means there are two isoforms normally expressed of their gene, while in significant DAS exons, there are no such trend of clusteration.

3. Third, I checked the PSI trend upon different AS types. I drawn boxplot using 3 datasets: All the detected events, Significant DAS Events (FDR<=0.01) and Top Significant Events (FDR<=0.01 & dPSI>=0.3).

Generally, Alternative Splicing Events undergo similar trend in M1 (two figures above) and M2 (figures on next page) macarphages. In all detected DASEs, monocytes and M1/M2 macrophages carry same trend: all AS types except MXE clustered in a higher inclusion part, especially with CA which has a vast majority around PSI 1. In the Significant DASEs (Mo/M1 has 4345, and Mo/M2 has 3704), the inclusiong ratio is lower, with more lower in macrophages. While in the Top significant DASEs, which has 185 events in Mo/M1 comparation and 152 in Mo/M2 comparation, there is a clear pattern difference between monocyte and M1/M2 macrophages. These events generally have median inclusion ratio in monocytes, while their inclusion ratio clustered in different part in macrophages, with AA/AD/IR clustered in lower PSI area, while CA moved to higher PSI area.

4. Next, I checked the AS types counts in the significant DASEs. Generally M1 and M2 macrophages share the same trend. So here I just show data of Monocytes/M1-macropahge. As expected, the most event type is casette exon (CA), followed by MXE. The ratio of AA, AD, IR increased in the top events.

5. At last, Heatmaps of the top 20 DASEs in the two datasets are shown.