Question about interpreting PCA values in HiC data

Question

[srhic]

Hello,

I have a question about PCA analysis specifically as it relates to HiC data. A common approach in HiC analysis is perform a PCA analysis on interaction matrices and classify genomic regions as active or repressed (A & B compartments) based on their Principal component 1 value. Any region having a negative PC1 value being repressed and .positive PC1 value being active.

I have done this analysis for my control and treatment conditions but am unsure about how to interpret them. If I am interested in a set of genomic regions, can I compare their PC1 value between conditions? For example, if the average PC1 value for my regions of interest in three different samples are 2, 0 and -1 respectively, does this mean that the phenotype (or A/B compartment status) of sample 2 is in a way intermediate between sample 1 and 3?

In other words does it make sense to use PC1 values for some sort of quantitative analysis, for example, by comparing mean PC1 values between conditions using a statistical test?

Thanks

Answer 1

[Kevin Blighe]

I have not worked on HiC, but have worked a lot on PCA in different contexts. How much variation is explained by PC1?

For example, if the average PC1 value for my regions of interest in three different samples are 2, 0 and -1 respectively, does this mean that the phenotype (or A/B compartment status) of sample 2 is in a way intermediate between sample 1 and 3?

From everything that I know about PCA, one cannot make a statement like this without probing further. How is the program that you're using conducting PCA, do you know? ...and it must be performing some transformation prior to performing PCA, correct? If you can share the exact functions that you are using, i would be happy to guide further...