update norm_line_to_residual and normalize_phospho description

[tobsecret]

Since we are using RidgeCV to normalize phospho, it is important to point out the vulnerability this approach has when the data is not normally distributed but instead bi-modal or multi-modally. Examples when this can happen is when two tumor types are combined such as lung and breast, or when tumor and normal data is analyzed together. In this example one would normalize phospho separately for lung tumors, lung normals, breast tumors and breast normals.

[granthussey]

I need to look into this, since we want to do pancan analysis

[tobsecret]

The pipeline already lets you do exactly this. You tell it which samples are part of which groups and it separates the data and normalizes it separately before combining it so you can do the correlation and clustering steps.