Modifications to PCA

• Print out eigenvalues as verification.
  • They represent the fraction of the variance each dimension explains. The first two should explain 90%, and thus suggests that a 2D plot is sufficient to show distances.
  • explained_variance_ratio_
• Decrease resolution of grid, then pool all data from all conditions in the grid in a big dataframe, for use with PCA.
• Use that to define the axes; have axes the same for all conditions.
• See which reactions dominate PC1 (probably glycolysis)
• Probably do not need standard scaling, but read documentation to make sure.