Literature-Based Discovery Predicts Antihistamines Are a Promising Repurposed Adjuvant Therapy for Parkinson's Disease

Abstract

Parkinson's disease (PD) is a movement disorder caused by a dopamine deficit in the brain. Current therapies primarily focus on dopamine modulators or replacements, such as levodopa. Although dopamine replacement can help alleviate PD symptoms, therapies targeting the underlying neurodegenerative process are limited. The study objective was to use artificial intelligence to rank the most promising repurposed drug candidates for PD. Natural language processing (NLP) techniques were used to extract text relationships from 33+ million biomedical journal articles from PubMed and map relationships between genes, proteins, drugs, diseases, etc., into a knowledge graph. Cross-domain text mining, hub network analysis, and unsupervised learning rank aggregation were performed in SemNet 2.0 to predict the most relevant drug candidates to levodopa and PD using relevance-based HeteSim scores. The top predicted adjuvant PD therapies included ebastine, an antihistamine for perennial allergic rhinitis; levocetirizine, another antihistamine; vancomycin, a powerful antibiotic; captopril, an angiotensin-converting enzyme (ACE) inhibitor; and neramexane, an N-methyl-D-aspartate (NMDA) receptor agonist. Cross-domain text mining predicted that antihistamines exhibit the capacity to synergistically alleviate Parkinsonian symptoms when used with dopamine modulators like levodopa or levodopa-carbidopa. The relationship patterns among the identified adjuvant candidates suggest that the likely therapeutic mechanism(s) of action of antihistamines for combatting the multi-factorial PD pathology include counteracting oxidative stress, amending the balance of neurotransmitters, and decreasing the proliferation of inflammatory mediators. Finally, cross-domain text mining interestingly predicted a strong relationship between PD and liver disease.

Keywords: Parkinson’s disease; antihistamines; artificial intelligence; machine learning; movement disorders; repurposed drugs.

Figure 1

Sequence of simulation searches in SemNet. For each simulation layer, specific “hub nodes” were identified. Hub nodes have a predicted strong relationship (e.g., HeteSim score) with the target node or query. The identified hub nodes were subsequently used as targets for the next layer of searches. The synthesis of information from all the SemNet 2.0 simulation layers contributed to the identification of antihistamines as a promising group of repurposed PD drugs. a = piroxicam; b = leflunomide; c = loratadine; d = ebastine; e = levocetirizine; f = ebastine.

Figure 2

Representative source node results from SemNet 2.0 simulations with the target node “Levodopa”. The source node type used was “Clinical Drug” (CLND). HeteSim scores were normalized to enable comparison of nodes across multiple simulations.

Figure 3

Representative source node results from SemNet 2.0 simulations with target nodes of “Antihistamine” and “Dopamine”. The source node types were “Clinical Drug” (CLND), “Pharmacologic Substance” (PHSU), and “Therapeutic or Preventative Procedure” (TOPP). HeteSim scores were normalized to enable comparison of nodes across multiple simulations.

Figure 4

Representative source node results from SemNet 2.0 simulation with target nodes of “Antihistamine” and “Levodopa.” The source node type was “Clinical Drug” (CLND). HeteSim scores were normalized to enable comparison of nodes across multiple simulations. Notice that ebastine reoccurs as a source node.

Figure 5

Normalized HeteSim scores of returned source nodes selected as hubs using “Parkinson’s Disease” as a SemNet 2.0 simulation target node. HeteSim scores were normalized to enable comparison of nodes across multiple simulations. Color code represents source node type: GNGM (gene or genome), DSYN (disease or syndrome), or PHSU (pharmacologic substance). Descriptions of shown source nodes: “Hypomyelination” is hypomyelination within brainstem and spinal cord; “WH” is Werdnig Hoffmann paralysis; “Granuloma” is granuloma of intestine; “Langer” is Langer mesomelic dysplasia syndrome; “Infection” is infection in the elderly; “Renal” is high renal threshold for glucose; “Liver injury” is drug-induced liver injury.

Figure 6

Overview of the cross-domain text-mining method. Over 33+ million journal articles from PubMed are text mined. Relationships are extracted according to the Unified Medical Language System (UMLS) ontology to construct a large-scale knowledge graph in a recently developed cross-domain text-mining software called SemNet 2.0 [8]. Artificial intelligence methods mine relationship patterns to identify promising candidates using “levodopa” and “Parkinson’s Disease” as the primary target nodes for the initial series of searches. Specifically, unsupervised learning rank aggregation assigned a ranking to filter the most promising repurposed drugs for PD.

Figure 7

An example subgraph (>99.99% pruned) obtained by querying the large SemMedDB knowledge graph using cross-domain text mining in SemNet 2.0. UMLS node types included are “Pharmacologic Substance” (PHSU), “Amino Acid, Peptide, or Protein” (AAPP), and “Disease or Syndrome” (DSYN). Note the full, unpruned graph is too large to visualize and would be intractable to the human eye.