Hi all,
I am creating an heatmap from gene expression in the brain using the allen brain atlas https://mouse.brain-map.org/
In practice I do a cumulative sum of gene level expression for a list of genes which I then devide by the number of genes (so I have a mean gene level of expression), let's callit F(one picture below)). Now I want to normalize for the total gene expression to see if the results is specific.So I do the same mean gene expression heatmap but for all discorevered genes in the allen atlas (about 15.000) let's call it F0 (another picture below).
What is the most correct way to normalize my heatmap coming from a list of genes?
I obtain complete different results if I do a voxel/pixel-wise subtraction ( [F-F0]>0 ) or division ( F/F0 ). What are the underlying priciple when you choose either to subtract or divide?
I am not sure about your goals,
anyway, when you want to normalize the image, you should operate with all pixel values from it (not mean value). It means that image intensities range is stretched to some other "normal" range. You can do linear normalization according to :
IN = ( I - Min )* (newMax - newMin)/(Max-Min) + newMin
Svitlana Alkhimova I operate at pixel value, with with mean value I mean that the volume was created averaging a set of volumes
- Badges
- Science topic
- Similar topics
- Engineering
- Electrical Engineering
- Signal Processing
- Image Processing