Biological Relevance

gene networks reflect tissue and condition specific biology mavatar discovery co expression gene networks successfully capture meaningful biological relationships by quantifying pairwise correlations between gene expression profiles and integration of vast number of datasets and samples, we obtained networks that represent the coordinated activity of genes within and across tissues the resulting structures revealed clear biological organization, consistent with expected tissue specific patterns and functional groupings, supporting the utility of this approach for biological interpretation methodology to evaluate biological relevance and uncover characteristic gene interactions within dina networks, we applied a network comparison framework that identifies both conserved molecular processes and context specific functional processes across the networks this approach yields two complementary outcomes dina network core genes and network specific genes capturing the distinctive functional wiring of each network the analysis revealed both underlying common core biological functions and distinct, context specific functionalities reflecting specialized activity in the biological condition of interest for each dina context specific network, iterative subnetwork comparisons are performed against all the other networks using the deltacon method ( docid\ m3ywqj3arjsy4zaf7wthc ) biologically relevant deltacon null distributions are used as a background ensuring that the similarities and differences reflect true biological organization rather than structural noise (see docid 0zknxsqa6r3cuqdntt6w for more details) this analysis yields two complementary gene lists for each network comparison similarity genes representing conserved patterns and dissimilarity genes representing network specific wiring these lists are subsequently aggregated and filtered to produce two network specific outputs dina network core genes and dina network specific signature genes for biological validation analysis we grouped these genes into louvain communities based on the network of interest and performed over representation analysis (ora) per network community for reactome, go\ bp, go\ mf and go\ cc pathway databases through g\ profiler ( docid\ m3ywqj3arjsy4zaf7wthc ) this approach provides a statistically grounded way to pinpoint functional gene modules that are conserved or rewired between biological contexts, enabling interpretation of tissue specific or disease specific pathways of the networks within the platform, users have access to the global dina network core gene list and all the dina network specific signature gene lists results and conclusions analysis of subnetworks allowed us to identify modules (gene groups) that are either conserved across tissues or uniquely present in one tissue or condition (see docid\ m3ywqj3arjsy4zaf7wthc ) shared subnetworks are likely to correspond to fundamental cellular processes active across multiple biological contexts, whereas tissue or disease specific subnetworks represent specialized or perturbed functions here we present the biological validation results from the following networks docid\ m3ywqj3arjsy4zaf7wthc , docid\ m3ywqj3arjsy4zaf7wthc , docid\ m3ywqj3arjsy4zaf7wthc , docid\ m3ywqj3arjsy4zaf7wthc , docid\ m3ywqj3arjsy4zaf7wthc and docid\ m3ywqj3arjsy4zaf7wthc across tissues, conditions, and disease states, the modules frequently showed enrichment for broad functional categories such as extracellular matrix organization, mitochondrial and respiratory processes, translation and ribosome biogenesis, immune response, rna splicing regulation, and cell cycle or dna replication in the following sections, we focus specifically on modules exhibiting tissue specific functional enrichment, which provide insight into biological processes characteristic of each network pancreas – general based on the enrichment analysis, the pancreas network specific signature genes in module 9 showed strong association with pancreatic physiological functions (table 1) the top enriched terms within this module include protein digestion and absorption related terms, which represent core functional processes of the pancreas in contrast, the similarity tail enrichment (table 2) results highlight subnetworks related to metal ion homeostasis and detoxification, including metallothionein binding metals and response to metal ions, consistent with the pancreas’ function in metal metabolism and stress response regulation co occurrence of metal ion homeostasis within the same module as the pancreatic secretion and digestion terms may suggests that the pancreas specific functional programs are built upon a conserved regulatory backbone of metal ion metabolism and detoxification ( docid\ m3ywqj3arjsy4zaf7wthc ) table 1 top enrichment results for pancreas – general module 9 specific signature genes source term adjusted p value reac digestion 2 98e 06 reac digestion and absorption 7 36e 06 go\ mf serine type endopeptidase activity 9 81e 06 table 2 top enrichment results for pancreas – general module 9 genes, co expressed between multiple networks source term adjusted p value reac metallothioneins bind metals 1 93e 06 go\ bp cellular response to zinc ion 2 17e 06 reac response to metal ions 5 82e 06 go\ bp intracellular zinc ion homeostasis 1 77e 05 go\ bp detoxification of copper ion 3 34e 05 brain – general brain network module 6 exhibited exclusive dissimilarity tail enrichment for neural and glial developmental pathways (table 3), such as axon ensheathment, myelination, oligodendrocyte differentiation, and glial cell differentiation additional terms such as neurogenesis, gliogenesis, and midbrain/substantia nigra development further support a central nervous system–specific module the absence of similarity tail enrichment indicates that these processes are strongly tissue specific and reflect specialized subnetwork organization within the brain network module 12 showed limited similarity tail enrichment (table 4) but extensive dissimilarity tail enrichment for synaptic structure and function, such as vesicle mediated transport in synapse, synaptic/postsynaptic membranes, nmda receptor–associated pathways, neurexin/neuroligin interactions, and exocytosis (table 5) this pattern indicates a brain specific synaptic module whose subnetwork organization is distinct in brain compared with other networks table 3 top enrichment results for brain – general module 6 specific signature genes source term adjusted p value go\ bp axon ensheathment 1 43e 19 go\ bp ensheathment of neurons 1 43e 19 go\ bp myelination 5 35e 18 go\ bp oligodendrocyte differentiation 1 81e 15 go\ bp glial cell differentiation 4 05e 14 go\ bp central nervous system myelination 1 28e 13 go\ bp substantia nigra development 9 41e 04 table 4 enrichment results for brain – general module 12 genes, co expressed between multiple networks source term adjusted p value go\ cc presynaptic active zone 1 55e 02 go\ bp developmental maturation 2 46e 02 table 5 top enrichment results for brain – general module 12 specific signature genes source term adjusted p value go\ bp vesicle mediated transport in synapse 3 58e 07 go\ cc synaptic membrane 5 76e 07 reac neuronal system 1 05e 04 go\ cc postsynaptic specialization 5 02e 04 go\ cc neuron to neuron synapse 7 19e 04 reac activation of nmda receptors and postsynaptic events 1 21e 03 skin – general skin network module 1 showed striking dissimilarity tail enrichment for epidermis and skin development, keratinocyte differentiation, keratinization, and cornified envelope formation, reflecting canonical skin barrier and keratinocyte functionalities (table 6) the corresponding similarity tail was enriched for broad vesicle membrane processes (table 7), consistent with conserved cellular regulation skin cells use endocytosis for basic functions such as taking in nutrients and signaling molecules which are closely connected to epidermis development these findings indicate that this module captures a skin specific functional program built on a shared signaling framework module 2 showed exclusive dissimilarity tail enrichment for melanocyte and pigmentation pathways, such as melanosome, melanosome membrane, melanin biosynthesis/metabolism, and mitf dependent programs (table 8), with no significant terms in the similarity tail we therefore interpret this as a skin specific pigmentation module skin module 16 displayed strong dissimilarity tail enrichment for keratin filament organization, and epidermal development, including intermediate filament cytoskeleton organization, cornified envelope formation, keratinocyte differentiation”, and hair follicle morphogenesis (table 9) the similarity tail contained no significant terms this pattern identifies a skin specific structural module, representing keratinocyte and epidermal differentiation programs characteristic of stratified epithelial tissue table 6 top enrichment results for skin – general module 1 specific signature genes source term adjusted p value go\ bp epidermis development 4 72e 13 go\ bp skin development 9 07e 10 go\ bp establishment of skin barrier 7 00e 08 go\ bp epidermal cell differentiation 1 19e 06 reac differentiation of keratinocytes in interfollicular epidermis in mammalian skin 3 60e 06 table 7 enrichment results for skin – general module 1 genes, co expressed between multiple networks source term adjusted p value go\ cc endocytic vesicle membrane 2 03e 02 table 8 top enrichment results for skin – general module 2 specific signature genes source term adjusted p value go\ cc melanosome membrane 6 05e 08 go\ cc chitosome 6 05e 08 go\ cc pigment granule membrane 6 05e 08 go\ bp pigment metabolic process 7 20e 07 go\ bp pigment biosynthetic process 2 74e 05 go\ cc melanosome 4 43e 05 go\ bp melanin biosynthetic process 3 47e 04 table 9 top enrichment results for skin – general module 16 specific signature genes source term adjusted p value go\ cc intermediate filament 1 28e 45 go\ cc keratin filament 2 30e 45 go\ cc intermediate filament cytoskeleton 7 36e 44 reac keratinization 4 28e 42 go\ bp intermediate filament organization 8 80e 12 go\ bp intermediate filament cytoskeleton organization 7 19e 11 go\ bp hair cycle 6 97e 10 liver – general liver network module 2 showed dissimilarity tail enrichment for a combination characteristic of hepatic biology, such as complement and coagulation cascades, acute phase response, γ carboxylation of protein precursors, and peroxisomal/fatty acid and amino acid metabolism (table 10) the corresponding similarity tail enrichment (table 11) highlighted general metabolic processes, such as small molecule and amino acid catabolism, nicotinamide/nucleotide metabolism, and metal ion response, consistent with conserved cellular functions together, these results support this module as liver specific on the dissimilarity side, built on a conserved metabolic backbone table 10 top enrichment results for liver – general module 2 specific signature genes source term adjusted p value go\ cc blood microparticle 1 15e 11 go\ mf serine type endopeptidase activity 3 72e 08 reac intrinsic pathway of fibrin clot formation 1 01e 04 reac complement cascade 1 11e 04 reac formation of fibrin clot (clotting cascade) 9 14e 04 table 11 top enrichment results for liver – general module 2 genes, co expressed between multiple networks source term adjusted p value go\ bp carboxylic acid catabolic process 1 12e 03 go\ bp organic acid catabolic process 1 12e 03 reac metabolism of amino acids and derivatives 1 52e 03 go\ bp pyridine nucleotide metabolic process 7 78e 03 go\ bp nicotinamide nucleotide metabolic process 7 78e 03 reac metallothioneins bind metals 7 87e 03 breast – breast cancer/tissue the breast cancer network exhibited several coherent and biologically interpretable modules the similarity tail of module 4 could be linked to endoplasmic reticulum protein folding, unfolded protein response, rrna processing, and ribosome biogenesis (table 12), marking a shared secretory, biosynthetic core typical of highly secretory epithelial tumors its dissimilarity tail emphasized transcriptional and rna processing regulation, such as major mrna splicing, rna polymerase ii elongation/pause release, and wdr5/nsl histone acetyltransferase complexes (table 13), reflecting cohort specific tuning of epigenetic and transcriptional output together these patterns indicate that breast cancer networks capture conserved cell cycle, biosynthetic, and stress response machinery, while the observed dissimilarity enrichments point to meaningful variation in dna repair, telomere biology, and transcriptional control breast cancer module 26 similarity tail was characterized by metal ion homeostasis functions, such as metallothionein and zinc/copper detoxification processes, a common oxidative stress signature (table 14) its dissimilarity tail highlighted telomere maintenance and telomerase associated repair, indicating heterogeneity in replicative lifespan control (table 15) the similarity tail of a dominant cell cycle/replication module (module 27) captured chromosome segregation, mitotic checkpoints, and dna replication, reflecting a conserved proliferative program across tumors (table 16) the associated dissimilarity signal was enriched for tp53 regulation, checkpoint control, and dna damage repair pathways, such as homologous recombination, brca/rad51, and base excision repair (table 17) this was consistent with variation in replicative stress and dna repair proficiency among cohorts table 12 top enrichment results for breast – breast cancer/tissue module 4 genes, co expressed between multiple networks source term adjusted p value go\ cc endoplasmic reticulum protein containing complex 1 06e 15 go\ bp ribosome biogenesis 3 60e 14 go\ bp rrna processing 6 64e 11 go\ bp protein folding 6 02e 08 table 13 top enrichment results for breast – breast cancer/tissue module 4 specific signature genes source term adjusted p value reac mrna splicing – major pathway 4 24e 07 reac processing of capped intron containing pre mrna 9 30e 07 reac epigenetic regulation of gene expression 2 70e 06 go\ bp positive regulation of gene expression, epigenetic 8 70e 06 go\ cc spliceosomal complex 8 11e 04 table 14 top enrichment results for breast – breast cancer/tissue module 26 genes, co expressed between multiple networks source term adjusted p value reac metallothioneins bind metals 2 96e 12 reac response to metal ions 1 28e 11 go\ bp detoxification of copper ion 6 79e 11 go\ bp intracellular zinc ion homeostasis 9 37e 09 go\ bp response to toxic substance 2 55e 04 table 15 top enrichment results for breast – breast cancer/tissue module 26 specific signature genes source term adjusted p value reac polymerase switching on the c strand of the telomere 3 45e 05 reac telomere c strand (lagging strand) synthesis 1 06e 04 go\ cc telomere cap complex 2 98e 04 reac chromosome maintenance 1 21e 03 go\ mf telomeric dna binding 1 47e 02 table 16 top enrichment results for breast – breast cancer/tissue module 8 genes, co expressed between multiple networks source term adjusted p value go\ bp chromosome segregation 7 42e 94 go\ bp nuclear division 5 43e 88 go\ bp organelle fission 2 89e 84 go\ bp nuclear chromosome segregation 1 34e 79 go\ bp dna replication 2 30e 58 table 17 top enrichment results for breast – breast cancer/tissue module 8 specific signature genes source term adjusted p value reac cell cycle checkpoints 3 32e 11 reac g2/m checkpoints 3 22e 10 reac dna repair 1 38e 07 reac regulation of tp53 activity 5 64e 07 reac impaired brca2 binding to rad51 4 42e 05 pbmc – general immune related modules appeared across multiple networks, reflecting conserved co expression structures underlying innate and adaptive immunity in pbmcs, this took the form of a stable antigen presentation/mhc ii module 1, which similarity tail showed baseline immune composition of blood mononuclear cells (table 18) while similar immune wiring may appear in other contexts, the pbmc module was distinguished by minimal activation signatures in its dissimilarity tail (table 19), suggesting maintenance of immune identity rather than context specific inflammatory activation in addition to pbmc module 1, two related immune modules were identified the first module captured innate effector and antigen processing functions, including neutrophil degranulation, phagocytosis, vesicle lumen, and lysosomal organization, while the corresponding dissimilarity enrichments highlighted toll like receptor and myd88 dependent signaling the second module represented broad cytokine signaling processes, encompassing tnf, il 17, il 10, and nf κb pathways, together with downstream transcriptional regulators these patterns reflect the mixed cellular composition and variable activation states characteristic of peripheral blood, indicating that the framework recovers coherent immune modules spanning both basal myeloid functions and cytokine mediated responses table 18 top enrichment results for pbmc – general module 1 genes, co expressed between multiple networks source term adjusted p value go\ cc mhc class ii protein complex 4 01e 23 go\ cc mhc protein complex 2 04e 20 go\ bp antigen processing and presentation of peptide or polysaccharide antigen via mhc class ii 4 19e 19 go\ cc lysosomal membrane 9 76e 16 table 19 enrichment results for pbmc – general module 1 specific signature genes source term adjusted p value go\ cc cell cortex 6 82e 04 reac maturation of spike protein 3 50e 03 reac translation of structural proteins 1 98e 02 reac calnexin/calreticulin cycle 4 15e 02 reac late sars cov 2 infection events 4 44e 02 summary our investigations confirm that mavatar discovery networks represent biologically coherent structures across all comparisons, the modules identified in both similarity and dissimilarity tails align with well established tissue and disease biology shared subnetworks capture conserved cellular and metabolic functions, while unique modules reflect context specific biology, such as immune regulation, tissue specialization, or microenvironmental interactions the consistency of these findings across multiple tissues and orientations demonstrates that the networks faithfully represent the underlying biological systems and capture both common and specialized processes relevant to each context in mavatar discovery, these results have been made accessible for graph construction the “network specific gene lists” represent genes with local connectivity patterns that differ from other networks, defining the functional rewiring unique to each context in contrast, the “dina network core genes” comprise one list of genes with connectivity profiles closely matching those in other networks, representing the shared cellular scaffold common to many systems references d koutra, j t vogelsteiny, c faloutsos, deltacon a principled massive graph similarity function proceedings of the 2013 siam international conference on data mining, sdm 2013 , 1304–4657 (2013) https //doi org/10 1145/2824443 l kolberg, u raudvere, i kuzmin, p adler, j vilo, h peterson, g\ profiler interoperable web service for functional enrichment analysis and gene identifier mapping (2023 update) nucleic acids res 51, w207–w212 (2023) https //pubmed ncbi nlm nih gov/37144459 e y chen, c m tan, y kou, q duan, z wang, g v meirelles, n r clark, a ma’ayan, enrichr interactive and collaborative html5 gene list enrichment analysis tool bmc bioinformatics 14, 128 (2013)