Analysis from the results of a label free mass spectrometry experiment on paraffin embedded samples stratified as follows:
The coverage analysis result was conducted here on both the first batch of samples, and the summary statistics of the coverage score for each sample is shown on the table below.
| Fib1 | Fib2 | Fib3 | Fib4 | RBC1 | RBC2 | RBC3 | |
|---|---|---|---|---|---|---|---|
| Min. | 0.23000 | 0.23000 | 0.19000 | 0.27000 | 0.05000 | 0.17000 | 0.19000 |
| 1st Qu. | 4.37000 | 5.18500 | 4.81500 | 4.75500 | 5.17500 | 5.26500 | 6.11500 |
| Median | 12.94500 | 11.45500 | 11.83000 | 12.30000 | 12.40000 | 11.70000 | 12.59000 |
| Mean | 18.50433 | 16.80495 | 17.79908 | 17.26768 | 16.64189 | 16.17239 | 17.05987 |
| 3rd Qu. | 26.50000 | 23.47250 | 25.30500 | 22.75500 | 23.10000 | 22.35000 | 23.66750 |
| Max. | 100.00000 | 91.55000 | 100.00000 | 93.50000 | 100.00000 | 100.00000 | 100.00000 |
The Venn Diagram below was generated using the first data set show that 145 proteins are overlapping between the RBC and FIB Groups - and this intersection is used for comparison between groups, 240 proteins are unique to the RBC samples, and 95 proteins are unique to the FIB samples.
Using the correlation matrix above to extract information on the abundancy profile of the top 50 highly correlated proteins with a threshold of +/- 0.80, the network analysis was carried out and resulted in the graph below. The node size is proportional with the degree (how connected the protein is), the red edges represent positive correlation and blue edges represent negative correlation. The optimal community structure was calculated for the graph, in using the maximal modularity score. This analysis resulted in the communities represented as the node colours and shown in the table below.
The same method for constructing the co-expression network used on the RBC samples was applied to the FIB dataset here, in this case using the 95 proteins unique to FIB samples from the first batch. The optimal community analysis resulted in the communities listed in the table below.
Here I used a linear model approach to assess differential abundance/expression between the two groups - this analysis resulted in 58 differentially abundant proteins. The tables below show the metrics of the top-ranked proteins from the linear model fit. The data was pre-processed with log-transformation and quantile normalisation to ensure that the expression distributions of each sample are similar across the entire experiment and not skewed.
The differentially abundant protein scores were used to perform the coexpression analysis on the comparison between groups. The heatmap below show differentially abundant proteins and resulted in clusters that are correspondent with the sample groups.
Pathway enrichment analysis were conducted on differentially abundant proteins related RBC and FIB using the InterMineR R package. Here the pathways were tested for over-representation in each of the proteins with fold change related with RBC and FIB relative to what is expected by chance and a p-value is computed for each pathway. The plots below represent the top 10 enriched pathways for the aforementioned communities - you can hover the bars for p-value information.