Search

Additional file 1: Fig. S1 Intra-PS similarity. The figure reports the distribution of the number of PS among the high-level DO classes (grey bars) as well as the average coefficient of intra-PS similarity (as mean ± SD; black squares). Interestingly, the PS belonging to DO classes that are well characterized in anatomical terms (e.g., disorders of the respiratory, reproductive and cardiovascular system) have intra-PS similarity scores above the mean similarity value of all the cumulative DO classes (dotted line). In contrast, the PS that are not anatomically defined (for instance, metabolic, infectious, hematologic, psychiatric and endocrine diseases) have intra-PS score lower than the average score. Fig. S2 Inter-PS similarity. The figure reports a selection of D that belong to the DO-classes DO:0014667 (Disease of metabolism; left side) and DO:0000863 (Nervous system disease; right side) and that are more similar to a D in another PS (in these examples, a metabolic D has the highest similarity with the indicated neurological D). The figure is derived from the DDSN-C. Fig. S3 Distribution of the P(k) in the bipartite D-DP graph. The distribution of connectivity P(k) approximates a power-law for the nodes indicating both the D (kD; top) and the DP (kDP; bottom). Fig. S4 Raising the thresholds and network fragmentation. Raising the threshold (w*) in the DDSN-C (A) and in the DP (B) progressively reduces the fraction of D nodes (white diamonds), DP nodes (gray diamonds), D−D edges (black squares) and D-DP edges (gray squares). Results are shown as percentage of the total number of nodes and edges in the whole networks (i.e., at a threshold of zero). The vertical dashed line indicates the threshold of 0.45 discussed in the text. The vertical dotted lines indicate the threshold of 0.80 applied to display the sub-networks of the DDSN-C (Fig. S5) and of the D-DP (Fig. S6). Fig. S5 A subnetwork of the DDSN-C. In the DDSN-C, applying a threshold for the average IC of the shared DP (w* > 0.80) causes loss of nodes and edges (as reported in Fig. S4 panel A), as well as fragmentation of the DDSN-C. The figure shows the remaining giant connected component, which contains 178 D nodes and 2379 D-D edges. The graph can be best visualized on a computer screen at high magnification. Fig. S6 A sub-network of the D-DP bipartite graph. In the D-DP bipartite graph, applying a threshold for the IC of the DP (w* > 0.80) causes loss of nodes and edges (as reported in Fig. S4 panel B), as well as fragmentation of the D-DP graph. The figure shows the remaining giant connected component, which contains 655 nodes (218 D and 437 DP) and 744 D-DP edges. White circles and gray rectangles are D and DP, respectively.

Additional file 1. Supplemental material.

Additional file 2. Explanation of the algorithm used to generate the similarity networks

Additional file 3: Fig. S1. Network fragmentation. Increasing the similarity threshold in the CSN (A), BSN-BP (B), BSN-CC (C), BSN-MF (D) and the general BSN (E) progressively reduces the fraction of nodes (open diamonds) and edges (closed diamonds), with a consequent fragmentation of the initial network. Results are shown as fractions of the total numbers of nodes and edges in the whole networks (i.e., at a similarity threshold of zero). Also shown is the threshold of 1.0 (vertical dotted lines) used for network analysis and the thresholds (vertical dashed lines) used to retain 20% of the initial nodes (horizontal dotted lines) and to display the networks in Figs. 2 (CSN), S2 (BSN-BP), S3 (BSN-CC), S4 (BSN-MF) and 3 (general BSN). Fig. S2. The BSN-BP. The BSN-BP, shown at a threshold of 2.5, contains 67 nodes linked by 135 edges and is fragmented into 14 islands and 9 clusters. In Figs. S2, S3 and S4, node color indicates the DO class (see inset of Fig. 2), while edge thickness is proportional to the weight w (i.e., the degree of HPO-based clinical similarity between PS). Fig. S3. The BSN-CC. The BSN-CC, shown at a threshold of 1.9, contains 63 nodes linked by 93 edges and is fragmented into 9 islands and 8 clusters. Fig. S4. The BSN-MF. The BSN-MF, shown at a threshold of 2.2, contains 58 nodes linked by 154 edges and is fragmented into 11 islands and 9 clusters.