DINA Network Similarity Validations

3 min

Validation for P-Value Calculation and Null Distributions

The validation analysis described under Biological RelevanceBiological Relevance incorporates DeltaCon similarities and their corresponding P-Values within the pipeline. Therefore, they can be considered as part of the DeltaCon similarity P-Value validation. To further ensure that our implementation of DeltaCon and null distributions give expected relative similarities between comparisons, and that the P-Values are significant where expected, comparative analyses of a known function across different subnetworks were performed. We analyzed subnetworks centered on the ZAP70 gene, which is associated with T-cell development and lymphocyte activation, and compared these subnetworks across multiple tissues.

CD8-Anchored (ZAP70) DeltaCon Validation Across Tissues

To evaluate the performance of the DeltaCon-based network similarity method, we compared gene subnetworks derived from multiple tissue-specific networks, namely: Multi-tissue – CD8+ T Cells, Multi-tissue – CD4+ T Cells, Blood – General, PBMC – General, Lung – General, Breast – General, Bone marrow – General, and Pancreas – General. The objective was to verify that biologically related networks exhibit higher similarity, while unrelated networks show lower similarity, with statistically significant P-Values in the expected comparisons. In this validation, the Multi-tissue – CD8+ T Cells network served as the reference (anchor), constructed around genes associated with T Lymphocytes receptor signaling and activation, centered on ZAP70. The hypothesis is that network similarity to the CD8 + T Lymphocytes reference should decrease with increasing biological dissimilarity. A detailed description of the expected relations between the networks is given in Table 1.

Table 1. Expected similarities between the networks from different tissues and the CD8+ T Lymphocytes as reference.

Compared to Multi-tissue – CD8+ T Cells	Expected Similarity	Rationale
Multi-Tissue – CD4+ T Cells	Very High	Closely related T-cell lineage; shared signaling (e.g., TCR/ZAP70), many co-expressed immune genes.
Blood – General	High	Mixed compartment that includes circulating CD8+ T lymphocytes.
PBMC – General	High	Enriched for lymphocytes (including CD8+), but still a mixture of cell types.
Lung – General	Moderate	Contains tissue-resident or infiltrating CD8+ T lymphocytes, especially during inflammation or in tumors.
Bone Marrow – General	Low–Moderate	Fewer mature CD8+ T lymphocytes; presence of precursors and diverse hematopoietic cells dilutes the T-cell signal.
Breast – General	Low–Moderate	Limited CD8+ infiltration, typically in tumor-adjacent or immune-active regions only.
Pancreas – General	Low	Generally low immune infiltration, particularly of CD8+ T lymphocytes, in non-inflamed tissue.

Observed DeltaCon similarities follow the expected biological gradient: CD4+ T Lymphocytes show the highest similarity to CD8+ T Lymphocytes (0.0385; p = 0.0216), followed by Blood (0.0258; p = 0.0905) and PBMC (0.0249; p = 0.1042) (Table 2). More dissimilar tissues show progressively lower similarity—Bone marrow (0.0194), Lung (0.0190), Breast (0.0152), and Pancreas (0.0139)—all non-significant (p > 0.05). Overall, the results match the hypothesis: biologically related networks are more similar to CD8+ T Lymphocytes, and unrelated tissues are less similar, with significance where expected (CD4+ T Lymphocytes).

Table 2. Observed Similarities Between the Networks from Different Tissues and the CD8+ T Lymphocytes as Reference.

Multi-tissue – CD8+ T Cells vs	Size	Number of Edges	Average Degree	Observed DeltaCon Similarity	P-Value	–log10 P-Value
Multi-Tissue – CD4+ T Cells	20	54	5.4	0.0385	0.0214	1.666
Blood – General	20	39	3.9	0.0258	0.0905	1.043
PBMC – General	20	41	4.1	0.0249	0.1042	0.982
Bone Marrow – General	20	31	3.1	0.0194	0.3072	0.513
Lung – General	20	25	2.5	0.0110	0.3398	0.469
Breast – General	20	15	1,5	0,0152	0,7596	0,119
Pancreas – General	20	8	0,8	0,0139	0,8727	0,059