Skip to content

Latest commit

 

History

History
36 lines (23 loc) · 2.9 KB

eMAGMA Tutorial PART_2.md

File metadata and controls

36 lines (23 loc) · 2.9 KB

PART 2

eMAGMA GENE-SET ANALYSIS

In this step we use the output of the gene-based analysis for Amygdala, to test which modules within a co-expression network are more highly associated with MDD risk. The test is carried out using the gene-sets function in MAGMA.

The co-expression network files are provided in the folder network_files.zip
https://github.com/AngelaMinaVargas/eMAGMA-tutorial/blob/master/network_files.zip

Details of the assembly of the co-expression networks are provided in Gerring et al., 2019a. Each co-expression network is subdivided into “modules” of highly correlated genes. Each co-expression network file has 2 columns: the modules are represented in the first column and are labelled as a colour; this is just a tag to identify the module or group during the analysis. The second column correspond to the genes ID’s.

As a first step, unzip the networks.zip file and run gene-set analysis using the code below:

./magma --gene-results Amygdala_emagma.genes.raw 
--set-annot network_files/Brain_Amygdala.txt col=2,1 
--out Amygdala_emagma 

The log file [Amygdala_emagma.log] shows that 24 gene-set (modules) were read from the input file, of these, 23 gene-sets include genes defined in the genotype data (containing a total of 1188 unique genes).

The results of the gene-set analysis are presented in the gsa.out file [Amygdala_emagma.gsa.out ]. This file shows that the smallest P-value=0.03703 is for the module: white, formed by 16 genes. Below is an example of the format of the output file.

# MEAN_SAMPLE_SIZE = 64134
# TOTAL_GENES = 1258
# TEST_DIRECTION = one-sided, positive (set), two-sided (covar)
# CONDITIONED_INTERNAL = gene size, gene density, sample size, inverse mac, log(gene size), log(gene density), log(sample size), 
VARIABLE            TYPE  NGENES         BETA     BETA_STD           SE            P
black                SET      55    0.0073145    0.0014962      0.13874      0.47898
skyblue3             SET       3     -0.64108    -0.031281      0.57506      0.86742
turquoise            SET     280     0.071266     0.029657     0.067712      0.14639
white                SET      16      0.45136     0.050599      0.25247      0.03703

After correcting for multiple testing, if significant genes are found, MAGMA generates a genes.sets.out file, with results for the significant sets. In this instance, modules in the Amygdala network are not significant. Gerring et al.,2009a uses a larger dataset (i.e. including results from 23andMe) and presents an example where significant associations between co-expression modules and MD are detected.

The user is invited to follow the steps in this tutorial to analyse their own data and make use of the annotation files and network files provided here. We encourage the user to explore the potential that these resources have, to find functional relationships and to mine GWAS summary data.