Results Interpretation

[Fnyasimi]

Thanks for this nice tool. From the tutorial you provide I have this results in cluster_network_1_scores_1.tsv file

# Observed cut height: 0.9583508
# Observed size of largest cluster at observed cut height: 7
# Expected size of largest cluster at observed cut height: 2.83
# Observed maximum ratio statistic: 2.473
# Expected maximum ratio statistic: 1.649
# p-value: 0.02
# Clusters:
i   j   k   l   n   o   p
a
b
c
d
e
f
g
h
m
q
r
s
t
u
v

Is there a page that explains the results in details. Am assuming each line after # represents a cluster. I would like to use this information to generate a dendogram and the network, what would be the best info to use for this? What would be the best criteria to find the highly altered subnetwork?

[TylerMclaughlin]

i can confirm my results are identical to this. It looks Hierarchical Hotnet found a single significant cluster of 7 genes, which would be the highly-altered subnetwork (with connections given in consensus_edges.tsv).

I agree it would be great to know which file would be best to use for cluster visualization with a dendrogram. i'm guessing it is the intermediate/network_1/similarity_matrix.h5 file? From that you can probably visualize the network, too.

[TylerMclaughlin]

Ah, so the similarity_matrix.h5 is just the toplogical similarity matrix. The vertex weights (mutation scores) are incorporated downstream of this. For the dendrogram linkages with SCCs (strongly connected components), use intermediate/network_1_scores_1/hierarchy_edge_list_0.tsv and intermediate/network_1_scores_1/hierarchy_index_gene_0.tsv.

The combined similarity matrix (with scores propagated along the interaction network) is not actually saved. It's generated in construct_hierarchy.py via combined_similarity_matrix(). It should be easy to change this code to save the propagated network for visualization or custom clustering.