# What statistics can be used for correlating gene expression on a specific organ with its function

My hypothesis is that a gene whose function is related to an organ will have higher gene expression in that organ compared to other organs.

An example of my hypothesis:

From a published literature, it is evident that gene x is function in liver lipid metabolism. So, as per my hypothesis, the expression of gene x would be higher in liver compared to other tissues or organs as,

genes       Kidney  Brain   Heart   Liver   Pancreas    Cornea
gene a      3.09    3.27    4.29    3.4       4.97       3.91
gene x      4.38    4.49    3.4     12.84     4.87       6.21
gene k      4.77    4.34    3.68    4.64      4.90       5.95
gene m      6.27    5.82    4.84    7.41      6.57       6.13
.
.
gene 20000  5.21   3.70     8.21    8.01      7.11       4.87


I have checked a few genes with reported organ related function and its level of gene expression in all organs. Most of the gene expression pattern justified my hypothesis.

What I have is

1) Normalized gene expression micro-array data for 20000 genes from 15 different organs as similar to the example above

2) A set of 35 genes for which function is identified in specific organs (based on literature published data).

I would like to know how can I apply statistics to test my hypothesis?

• You should 1. explain your hypothesis 2. explain what kind of data you have – Juho Kokkala Apr 26 '18 at 4:48

Unfortunately the most common test here is the dirty fold-change cutoffs.

gene x      4.38    4.49    3.4     12.84     4.87       6.21


For this gene, your Liver has 12.84 and your second largest is 6.21. Your log2 ratio will be log2(12.84/6.21).

Compare this cutoff with your preferred thresholds, very common in bioinformatics.