Frequent edge sets in human connectome graphs

By Mate Fellner

One way to examine the connections between the distinct parts of the human brain is the braingraph, or the connectome. The nodes of the connectome are not neurons, but distinct areas of the gray matter, which can be found in any average and healthy human brain, so we can compare whether two edges are present in the graphs between the same nodes. The connectomes are mapped from human brains by diffusion MR imaging.

In our work, we used the apriori algorithm to find frequent edge sets of the connectomes, then the frequent connected edge sets. For this we modified the method to increase efficiency for these special structures instead of selecting them from the previous result sets.

The average running time of the algorithm is exponential in the size of the frequent sets, so we searched for small sets with size less then 7 with 90% support, which means a set was considered frequent if 90% of the brains contained it. The following diagram shows the number of the frequent sets for different supports in logarithmic scale.


Next we separated the dataset to two samples, to male and female braingraphs and repeated the same process. Then we made a list of frequent connected sets which have at least 90% support in the female or the male sample and for every set we applied chi-squared test to find significant differences.

In the first picture there is a connected 6-set which is significantly more frequent in female brains, and in the second there is another, which is more frequent in male brains.


Mate Fellner is a first year Applied Mathematics MSc student at ELTE, he is working with the PIT Bioinformatics Group under the supervision of Vince Grolmusz, where they solve problems arising in life sciences by mathematical and informatical means.
%d bloggers like this: