So this week consisted of a clustering analysis of the tissue types in my analysis. Based on the Pearson correlation matrix I generated last week, I used Ward’s hierarchical clustering in order to group similar tissues together. My result ended up looking like this:
It’s not very aesthetically pleasing, but it does the job. Looking at this from just a biological standpoint, the clustering seems to make sense. Female reproductive organs group together based on their RNA-seq expression, and so do other tissues with related or similar functions. I compared this clustering to a larger and much more detailed clustered heatmap in the GTEx paper I replicated this method from, and the trends appeared to follow those in my figure. Since this was validated, I can start to find tissue-specific genes for tissues with and without abnormal samples. However, as my PI suggested, I might use PCA plots to analyze some of the tissue samples first.