Drug Repurposing and Network Pharmacology

Drug repurposing (also termed drug repositioning or therapeutic switching) is the systematic strategy of identifying, validating, and developing new clinical indications for existing FDA-approved, discontinued, or clinical-stage investigational pharmaceuticals outside the scope of their original medical hypothesis. Integrated with network pharmacology—a discipline formalized by Andrew L. Hopkins in 2007–2008 that replaces the reductionist "one-drug, one-target, one-disease" model with systems-level analysis of polypharmacology across biological networks—drug repurposing exploits the reality that most bioactive small molecules interact with an average of 6 to 11 macromolecular proteins within the human proteome, and that single molecular targets frequently regulate pleiotropic signaling pathways across distinct organ systems.

In contrast to de novo drug discovery—which requires 10 to 15 years, $1.3 billion to $2.6 billion in capitalized investment, and suffers a ~90% clinical attrition rate—a repurposed drug has already completed chemical synthesis scale-up (CMC), GLP animal preclinical toxicology and safety pharmacology, and Phase I human pharmacokinetics and tolerability profiling. Consequently, repurposed candidates can often bypass 3 to 6 years of early discovery and pre-IND qualification to enter directly into Phase II proof-of-concept clinical trial phases, frequently utilizing the hybrid Section 505(b)(2) NDA regulatory pathway.

Repurposing discoveries fall into two distinct mechanistic categories depending on whether the new therapeutic activity arises from the drug's known primary target or an unrecognized secondary protein target:

On-Target Repurposing (Disease-Centric Mechanism)

In on-target repurposing, a drug's established pharmacological mechanism on its primary target protein (T₁) is applied to a different pathophysiological condition once human genetics or clinical observation reveals that T₁ drives both diseases:

1. Sildenafil (PDE5 Inhibitor): Originally synthesized by Pfizer chemists as an anti-anginal agent (UK-92,480) to inhibit cyclic guanosine monophosphate (cGMP)-specific phosphodiesterase type 5 (PDE5) in coronary vascular smooth muscle. During Phase I clinical trials in Merthyr Tydfil, Wales, sildenafil showed modest anti-anginal activity but consistently induced penile erections by potentiating nitric oxide (NO)–cGMP smooth-muscle relaxation in the corpus cavernosum—leading to its approval for erectile dysfunction (Viagra, 1998) and subsequently for pulmonary arterial hypertension (Revatio, 2005), where PDE5 is abundantly expressed in pulmonary vasculature.
2. Minoxidil and Finasteride: Minoxidil, an oral ATP-sensitive potassium (K_ATP) channel opener developed for severe refractory hypertension, caused widespread hypertrichosis as a side effect and was reformulated as a topical treatment (Rogaine) for androgenetic alopecia. Similarly, finasteride, a competitive inhibitor of steroid 5α-reductase type II (which blocks conversion of testosterone to dihydrotestosterone [DHT]), was approved at 5 mg/day (Proscar) for benign prostatic hyperplasia and subsequently repurposed at 1 mg/day (Propecia) for male pattern hair loss.
3. Baricitinib (JAK1/2 and NAK Inhibitor): During the 2020 SARS-CoV-2 pandemic, knowledge-graph machine learning (BenevolentAI) predicted that the rheumatoid arthritis Janus kinase (JAK1/JAK2) inhibitor baricitinib also inhibits Numb-associated kinases (AAK1 and GAK) governing clathrin-mediated viral endocytosis while blunting IL-6 cytokine storm, leading to rapid randomized validation in the ACTT-2 and RECOVERY trials and FDA approval for hospitalized COVID-19.
4. GLP-1 Receptor Agonists: Originally developed as glucose-dependent insulinotropic agents for type 2 diabetes mellitus, glucagon-like peptide-1 (GLP-1) receptor agonists (such as semaglutide and tirzepatide) were repurposed via hypothalamic satiety signaling and systemic anti-inflammatory effects for chronic obesity, cardiovascular risk reduction, metabolic dysfunction-associated steatohepatitis (MASH), and heart failure with preserved ejection fraction (HFpEF).

Off-Target Polypharmacology Repurposing (Target-Centric Mechanism)

In off-target repurposing, a drug engages a previously unknown secondary macromolecular target (T₂) distinct from its canonical target (T₁):

1. Thalidomide: Withdrawn in 1961 for severe teratogenicity when used as a sedative, thalidomide was serendipitously found by Jacob Sheskin in 1965 to resolve erythema nodosum leprosum (ENL) via TNF-α suppression, and by Robert D'Amato and Judah Folkman in 1994 to inhibit angiogenesis, leading to its breakthrough efficacy in multiple myeloma. In 2010–2014, Hiroshi Handa, Benjamin Ebert, and William Kaelin revealed thalidomide's true off-target molecular mechanism: it acts as a molecular glue that binds the Cereblon (CRBN) E3 ubiquitin ligase substrate receptor, reprogramming CRBN to polyubiquitinate and degrade the lymphoid transcription factors IKZF1 (Ikaros) and IKZF3 (Aiolos) in myeloma cells (while degrading SALL4 in fetal limb buds).
2. Amantadine: Approved in 1966 as an antiviral targeting the influenza A M2 proton channel, amantadine was observed in 1969 to relieve tremor and levodopa-induced dyskinesia in Parkinson's disease through off-target non-competitive antagonism of striatal N-methyl-D-aspartate (NMDA) glutamate receptors.
3. Ketamine / Esketamine: Synthesized in 1962 as a dissociative general anesthetic blocking NMDA receptors, sub-anesthetic intravenous ketamine and intranasal (S)-ketamine (Spravato, 2019) were repurposed for treatment-resistant depression, triggering rapid glutamatergic AMPA receptor activation, brain-derived neurotrophic factor (BDNF) release, and mTOR-dependent prefrontal synaptogenesis within hours.

Interactome Topology and Network Proximity (z_c)

Classical pharmacology assumes that a disease can be corrected by selectively inhibiting a single mutated protein. However, human complex diseases (such as Alzheimer's disease, schizophrenia, atherosclerosis, and metastatic cancer) arise from perturbations across interconnected disease modules within the human protein–protein interaction (PPI) network (the interactome)—which comprises ~20,000 protein nodes linked by >300,000 physical and regulatory edges. Because biological networks exhibit scale-free topology (where degree distribution follows a power law P(k) ~ k^−γ), inhibiting a single peripheral node is frequently bypassed by redundant signaling pathways, whereas inhibiting a high-connectivity "hub" protein (such as pan-PI3K or pan-HDAC) causes systemic toxicity.

Network pharmacology seeks either rational polypharmacology (a single multi-target ligand or synergistic drug combination that modulates multiple nodes within a disease module simultaneously) or topological proximity between known drug targets and disease genes. As formulated by Albert-László Barabási and Joseph Loscalzo, given the set of a drug's protein targets T and the set of disease-associated proteins S mapped onto the human interactome, the closest shortest-path network distance d_c(S, T) is defined as:

d_c(S, T) = (1 / ||*T*||) ∑_{t ∈ T} min_{s ∈ S} d(s, t)

where d(s, t) is the shortest path length (number of PPI edges) between disease protein s and drug target t. By comparing d_c(S, T) against a degree-matched random reference distribution (with mean μ_d and standard deviation σ_d), the network proximity z-score is calculated as:

z_c = [ d_c(S, T) − μ_d ] / σ_d

Drugs exhibiting statistically significant network proximity (z_c < −1.5) to a disease module demonstrate substantially higher clinical enrichment and therapeutic efficacy in randomized trials and electronic health record (EHR) validation cohorts than drugs acting outside the disease neighborhood.

Transcriptomic Signature Reversal (Connectivity Map & LINCS L1000)

A complementary genome-wide computational repurposing engine is transcriptomic signature reversal, pioneered by Justin Lamb, Todd Golub, and colleagues at the Broad Institute via the Connectivity Map (CMap) and the NIH Library of Integrated Network-Based Cellular Signatures (LINCS L1000) database (which catalogs >1.3 million gene-expression profiles across human cell lines treated with >20,000 small molecules and genetic CRISPR/shRNA perturbations).

1. Disease Signature Extraction: Differential RNA-sequencing between diseased patient tissue and healthy control tissue defines a ranked disease signature (S_disease) of the top up-regulated and down-regulated genes.
2. Connectivity Enrichment Scoring: Using a non-parametric, rank-based Kolmogorov–Smirnov (KS) enrichment statistic (analogous to Gene Set Enrichment Analysis [GSEA]), the algorithm scans S_disease against every drug perturbation profile (S_drug) in the CMap/L1000 reference library.
3. Signature Reversal Selection: Drugs that yield a strongly negative connectivity score (τ → −100)—meaning the drug down-regulates the exact genes up-regulated by the disease and up-regulates the genes suppressed by the disease—are prioritized as repurposing candidates capable of restoring the diseased transcriptome toward homeostasis (a method that successfully identified cimetidine for lung adenocarcinoma and topiramate for inflammatory bowel disease).

Alongside computational in silico engines, physical high-throughput screening of annotated clinical libraries—most notably the Broad Institute Drug Repurposing Hub (~7,000 FDA-approved and clinical-phase compounds) and the Scripps ReFRAME (Repurposing, Focused Rescue, and Accelerated Medchem) library (~12,000 compounds that have reached clinical trials or significant preclinical characterization)—enables rapid phenotypic screening of human organoids and patient-derived cells.

Why do many repurposed drugs face commercial and patent challenges despite strong clinical science?

If an off-patent generic drug (such as generic metformin, aspirin, or dexamethasone) is proven effective in a new disease at its already marketed dosage strength, a sponsor obtains only a "method-of-use" patent and 3 years of 505(b)(2) regulatory exclusivity. Because physicians and pharmacists can legally prescribe and dispense low-cost generic tablets off-label for the new indication ("skinny labeling" under Section viii), sponsors frequently cannot recoup the $50 million to $150 million cost of Phase III trials unless they develop a novel formulation, prodrug, delivery device, or distinct dosage strength (as Merck did by pricing 1 mg finasteride [Propecia] separately from 5 mg finasteride [Proscar]).

What is the difference between Drug Repurposing and Selective Optimization of Side Activities (SOSA)?

Drug repurposing takes the exact same clinical molecule (without modifying its covalent chemical structure) and develops it for a new therapeutic indication. Conversely, Selective Optimization of Side Activities (SOSA), formulated by Camille Wermuth, uses an existing approved drug that exhibits weak off-target affinity for a secondary receptor as a medicinal-chemistry starting point; chemists then synthesize structural analogues to dial in potency at the new off-target receptor while dialing out the original primary pharmacological activity (for example, how antibacterial sulfonamides were chemically modified to create thiazide diuretics and sulfonylurea antidiabetics).

How does Electronic Health Record (EHR) pharmacoepidemiology validate computational repurposing predictions?

Before launching a costly Phase II/III randomized trial for a computationally predicted drug–disease pair (identified via CMap signature reversal or interactome network proximity z_c), investigators query longitudinal insurance claims and EHR databases containing millions of patient records (such as Optum, MarketScan, or the VA Corporate Data Warehouse). Using active-comparator, new-user cohort designs with propensity-score matching across hundreds of baseline covariates, researchers test whether patients prescribed Drug A for its original indication exhibit a statistically lower hazard ratio (HR < 1.0) of developing Disease B compared to matched patients prescribed a comparator drug.

1. Pushpakom, S., Iorio, F., Eyers, P. A., Escott, K. J., Hopper, S., Wells, A., ... & Pirmohamed, M. (2019). "Drug repurposing: Progress, challenges and recommendations." Nature Reviews Drug Discovery, 18(1), 41–58. https://doi.org/10.1038/nrd.2018.168
2. Hopkins, A. L. (2008). "Network pharmacology: The next paradigm in drug discovery." Nature Chemical Biology, 4(11), 682–690. https://doi.org/10.1038/nchembio.118
3. Lamb, J., Crawford, E. D., Peck, D., Modell, J. W., Blat, I. C., Wrobel, M. J., ... & Golub, T. R. (2006). "The Connectivity Map: Using gene-expression signatures to connect small molecules, genes, and disease." Science, 313(5795), 1929–1935. https://doi.org/10.1126/science.1132939
4. Guney, E., Menche, J., Vidal, M., & Barábasi, A. L. (2016). "Network-based in silico drug efficacy screening." Nature Communications, 7(1), 10331. https://doi.org/10.1038/ncomms10331