Human Interactome

The human interactome is the complete set of molecular interactions that occur within a human cell, tissue, or organism — most commonly the comprehensive map of protein-protein interactions (PPIs), but also encompassing protein-DNA, protein-RNA, RNA-RNA, and metabolic interactions. It is the molecular analogue of the neural connectome: not a list of parts, but a network whose structure constrains and enables biological function. The human interactome contains on the order of 10^5 to 10^6 protein-protein interactions among roughly 20,000 protein-coding genes, though estimates vary by experimental method and coverage.

The interactome is not merely a database. It is a constraint system: diseases do not typically arise from the failure of a single gene in isolation, but from the perturbation of interaction modules — subgraphs of the network in which mutations propagate through physical and regulatory links. This network perspective has transformed genetics from a parts-list discipline into a systems discipline, where the relevant unit of analysis is often the topological neighborhood of a gene rather than the gene itself.

Mapping the Interactome: Methods and Coverage

The human interactome has been mapped by two complementary strategies. High-throughput experimental methods — yeast two-hybrid screening, affinity purification mass spectrometry, and proximity labeling — generate interaction data at proteome scale, but with high false-positive rates and incomplete coverage. Each method captures a different biochemical reality: yeast two-hybrid detects binary physical interactions, while affinity purification detects co-complex membership. The intersection of multiple methods is more reliable than any single method, but the union is more complete — a trade-off between precision and recall that has no clean resolution.

Curated databases — BioGRID, STRING, IntAct — integrate experimental evidence with computational predictions (phylogenetic profiling, gene co-expression, domain-domain interaction patterns) to produce consolidated networks. These databases are indispensable for research but are themselves artifacts of curation bias: well-studied proteins and diseases are overrepresented, while poorly funded research areas leave large regions of the network dark. The interactome we have is not the interactome that exists; it is the interactome that funding and methodological fashion have made visible.

Disease Modules and Network Medicine

The central operational concept in interactome-based disease analysis is the disease module: a connected subgraph of the interactome within which disease-associated genes are significantly clustered. The disease module hypothesis, developed by Barabási and colleagues, states that if two disease genes interact directly or through a short path in the interactome, their corresponding diseases share phenotypes, comorbidities, or drug sensitivities. This hypothesis has been validated for cancers, cardiovascular diseases, and Mendelian disorders, though the strength of the effect varies with disease class.

The practical implication is that drug targets need not be disease genes themselves. A drug that perturbs a neighbor of a disease gene in the interactome — even a non-disease gene — can indirectly restore the network state disrupted by the disease mutation. This principle, called network proximity, underlies the network medicine research program and has motivated the repurposing of existing drugs whose primary targets are not obviously disease-related but occupy strategic positions in disease modules.

Structural Properties and Biological Meaning

The human interactome, like most biological networks, is a scale-free network with a degree distribution that follows a power law: most proteins participate in few interactions, while a small number of hub proteins participate in many. These hubs are enriched for essential genes — genes whose deletion is lethal — suggesting that topological centrality is a proxy for biological importance. However, the scale-free property has been challenged on statistical grounds, and the correlation between degree and essentiality is stronger for some interaction types (binary) than others (co-complex).

The interactome also exhibits modularity: densely connected subgraphs (protein complexes, signaling pathways, organelle-specific interactions) that are sparsely connected to the rest of the network. Modularity is thought to confer robustness by confining perturbations to local regions, though the evolutionary origin of interactome modularity remains debated. Some argue it reflects selection for functional separation; others argue it emerges generically from duplication-divergence dynamics, in which genes duplicate and their interaction partners diverge over evolutionary time, producing clusters even without selection for modularity.

The human interactome is frequently treated as a static map, a reference architecture against which disease mutations can be plotted. This is a category error. The interactome is a dynamic, context-dependent structure: protein interactions change with post-translational modification, cellular compartment, developmental stage, and environmental condition. The network medicine program risks repeating the reductionist mistake it claims to transcend — treating a time-averaged, cell-averaged snapshot of molecular interactions as the ground truth of biological function. What is needed is not a bigger static map but a theory of how interaction networks reconfigure under perturbation, and whether those reconfigurations themselves are the disease.