Methods

The ranking parameter, which is informative of the molecule relevance for the phenotype under study, may be used as a guide to scan for subnetwork enrichment through the entire list. This ensures us to avoid a threshold imposition which may a affect the final biological conclusions. With this purpose, a framework for subnetwork enrichment was designed. The framework proposed act in accordance with the steps below:

1. The ranked list of n molecules is subdivided into a sequence of additives partitions.
2. The proteins corresponding to any of the partitions are mapped onto the interactome scaffold and the minimum connected network (MCN) and the parameter of interest (the average nodes per component of the MCN) are found for them.
3. Then we seek for the most relevant partition as follows:
   • First, ordering the parameter of interest according to the ranked list, all relative maxima are identified. The partitions so selected represent situations where a new protein capable of connecting to the previous ones is added to the previous partitions.
   • Second, the score is computed. The score can be seen as a balance between the increase in connected nodes and the distance to the top of the ranked list.
   • Third we choose the best partition.

An empirical p-value is calculated as the proportion of 2000 random sublists of the same size as best partition (which corrects the size effect) with a value for the network parameter greater than the best partition.

Optionally, a list of seed molecules may be incorporated. Typically this list represents genes of interest like already known disease genes or phenotype-related genes that the program will include in the testing list. The selection procedure is the same than describe above, but keeping always the seed molecules within the list i.e. the ranked list of n molecules is subdivided into a sequence of additives partitions constrained to contain the seed list.