Network‐based prediction and real‐world patient data observation identifies Doxycycline and Irbesartan as candidate medicines for Alzheimer’s disease and Traumatic Brain Injury

Marina Bykova, Michael M. Danziger, Michal Rosen‐Zvi, Andrew A. Pieper, Feixiong ChengAlzheimer 2024 · DOI

TL;DR

This paper tackles the challenge of finding new treatments for Alzheimer's Disease (AD) and Traumatic Brain Injury (TBI) by combining network biology with real-world patient data. They identified doxycycline and irbesartan as promising drug candidates for AD, suggesting a novel approach to drug repurposing for these complex neurological conditions.

What Problem Does This Paper Address?

Alzheimer's Disease (AD) and Traumatic Brain Injury (TBI) are significant health burdens, and effective treatments are sorely lacking. TBI is a known risk factor for developing AD, highlighting a critical need for interventions that could address both conditions. Traditional drug discovery is a lengthy and expensive process. This paper aims to accelerate the identification of potential drug candidates by repurposing existing medications, specifically by integrating biological network information with large-scale patient health records.

Methods

The researchers employed a two-pronged approach:

  1. Network-Based Prediction: They started by compiling lists of genes associated with TBI (69 genes) and AD (144 genes) from existing genetic and multi-omics studies. These gene lists were then mapped onto a human protein-protein interactome network. Using a "network proximity analysis," they identified drugs that are likely to interact with these disease-associated genes, prioritizing those with a strong predicted connection. This method leverages the idea that drugs targeting proteins close to disease-related proteins in a biological network are more likely to be effective.
  2. Real-World Patient Data Analysis: To validate the predictions from the network analysis, the researchers turned to a massive real-world patient database (MarketScan, containing data from ~170 million individuals). They investigated two patient cohorts:
  3. A general cohort of patients over 70 years old.
  4. A specific cohort of patients who had experienced a traumatic brain, head, or facial injury. They looked for associations between the prescription of certain drugs and a reduced incidence of AD in these cohorts.

Datasets

  • Human Protein-Protein Interactome Network: This served as the biological network scaffold for the prediction analysis.
  • MarketScan Patient Database: A large-scale, real-world dataset of electronic health records (EHR) from approximately 170 million individuals. This dataset was used for the observational study to validate drug associations with AD incidence.

Key Insights and Ideas

  • Bridging the Gap: The study successfully demonstrates how integrating complex biological network information with large-scale, real-world patient data can be a powerful strategy for drug repurposing.
  • Doxycycline and Irbesartan as AD Candidates: The most significant finding is the identification of doxycycline (an antibiotic) and irbesartan (a cardiovascular drug) as potential candidates for treating AD. This is particularly interesting because these drugs are already approved for other conditions, potentially shortening the path to clinical use.
  • TBI as a Pathway to AD: The paper reinforces the link between TBI and AD, suggesting that interventions for TBI might also have protective effects against AD.
  • Network Proximity as a Predictive Tool: The network proximity analysis proved effective in prioritizing a manageable list of candidate drugs from a vast pool of possibilities.

Key Results

  • Candidate Drug Identification: The network proximity analysis identified 66 candidate drugs for AD and 53 for TBI, with a high statistical significance (z-score < -3, p-value < 0.01).
  • Doxycycline Association: In the over-70 cohort, doxycycline use was significantly associated with a reduced incidence of AD compared to other anti-infective and antibacterial drugs (OR = 0.92, adjusted 95% CI 0.88-0.97, p-value = 0.002).
  • Irbesartan Association: Irbesartan use was significantly associated with a reduced incidence of AD in the overall cohort, both when compared to patients taking general cardiovascular drugs (OR = 0.82, adjusted 95% CI 0.71-0.95, p-value = 0.002) and specifically to those taking renin-angiotensin agents (OR = 0.82, adjusted 95% CI 0.71-0.97, p-value = 0.002).

Limitations and Future Work

  • Observational Nature: The real-world data analysis is observational, meaning it shows an association but cannot definitively prove causation. While the results are statistically significant, they don't confirm that doxycycline or irbesartan prevent AD.
  • Need for Validation: The authors explicitly state that further functional and clinical validation is warranted. This means laboratory experiments to understand the biological mechanisms and, crucially, clinical trials in humans are needed to confirm these findings and establish a causal relationship.
  • Mechanism of Action: While the study identifies candidate drugs, the precise biological mechanisms by which doxycycline and irbesartan might influence AD risk are not detailed and require further investigation.

Summary updated

Abstract

Abstract Background Traumatic Brain Injury (TBI) is one of the most common nonheritable causes of Alzheimer’s disease (AD). However, there is lack of effective treatment for both AD and TBI. We posit that network‐based integration of multi‐omics and endophenotype disease module coupled with large real‐world patient data analysis of electronic health records (EHR) can help identify repurposable drug candidates for the treatment of TBI and AD. Method We have collected a total of 69 TBI‐related genes and 144 AD‐related genes derived from human genetic and multi‐omics findings. We conducted network proximity analysis to prioritize likely drug candidates under the human protein‐protein interactome model. Using the MarketScan patient database (∼170 million individuals), the drugs were further investigated for association with a reduced incidence of AD against two patient cohorts: A cohort comprised of all patients over 70, and a TBI cohort comprised of patients who have sustained a traumatic brain, head, or facial injury. Result Using network proximity‐based analysis of drug‐target networks and disease‐associated genes under the human protein‐protein interactome network, we identified 66 and 53 candidate drugs for AD and TBI, respectively (z‐score ≤ ‐3, p‐value ≤ 0.01). Drugs present within or related to drugs within both candidate lists, such as irbesartan and doxycycline, were especially prioritized for further real‐world patient data validation. In the subsequent Marketscan analysis, we found that doxycycline is significantly associated with a reduced incidence of AD in the over‐70 cohort as compared to cohort patients taking other anti‐infective and antibacterial drugs (OR = 0.92, adjusted 95% CI 0.88‐0.97, p‐value = 0.002). In addition, prescription of irbesartan was significantly associated with a reduced incidence of AD as compared to all cohort patients taking cardiovascular (OR = 0.82, adjusted 95% CI 0.71‐0.95, p‐value = 0.002) and renin‐angiotensin agents (OR = 0.82, adjusted 95% CI 0.71‐0.97, p‐value = 0.002). Conclusion We demonstrated that combining network‐based predictions and large real‐world patient data observations has identified doxycycline and irbesartan as candidate treatments for AD. Further functional and clinical validation is warranted to examine a causal relationship between doxycycline and irbesartan with a future risk of developing AD.
BibTeX
@inproceedings{bykova2024networkbased,
  author = {Marina Bykova and Michael M. Danziger and Michal Rosen‐Zvi and Andrew A. Pieper and Feixiong Cheng},
  title = {{Network‐based prediction and real‐world patient data observation identifies Doxycycline and Irbesartan as candidate medicines for Alzheimer’s disease and Traumatic Brain Injury}},
  booktitle = {Alzheimer},
  year = {2024},
  doi = {10.1002/alz.089712},
}
Figure