Best way to normalize transcript reads data for PCA & correlation and further biostat analysis

Question

[Faith]

Hello all,

I would like your advice on how to normalize, a Log2 normalized values of transcripts from DEGs flowerets in eight consecutive stages of development across, each stage has 3 biological replicates with their reads data.

I know they're logged, but my supervisor suggested we scale(scale=T, center=T) is this the best way to scale them? I tried it but the heatmap with scaling and without is nearly the same.

I'm planning to do PCA &Correlation on the data and further biostat analysis. Any advice is appreciated.

Answer 1

[Jeremy]

Scaling the data will give you Z-scores. This is a standard way of preparing data for PCA. I would also suggest trying a center-log-ratio transformation, which sometimes does a better job of separating the data for PCA. You can use the clr() function in the compositions package in R to do that.