I am analyzing the genomes of different closely related bacteria and I want to investigate the coverage of metabolic pathways based on genomic data.
I annotated the genomes with the RAST server and in R I created a matrix where the column headers are names of the bacteria and row names correspond to pathway names. In the matrix, I have the number of different genes mapped of each bacteria on each pathway.
I could now already make a heatmap but it would be quite useless because I have now absolute counts in my matrix and I can only make comparison by row but not compare the coverage of different pathways.
My question is, how can I normalize my data that it can be compared across the entire heatmap?
I also found this paper here from Verma et al. who did such a similar thing in Figure 4.
My second question is what are they doing there with Pearson correlation?
Further, the coding sequences were processed for functional annotation using the bi-directional best-hit (BBH) assignment method on KEGG Automatic Annotation Server (KAAS) [62]. This annotation was then used for biological family construction using protein family prediction on MinPath [63]. The top 50 subsystems were selected based on normalized values obtained by dividing with the lowest value for the genes in the respective pathways. Finally, the nine Sphingobium strains and enriched pathways were clustered heirarchially using Pearson correlation with 0.8% minimum abundance and a heat map was constructed in MeV4.9.0 [64].
I know this is a long question but any help would be highly appreciated. Thank you for taking your time for my questions.
