Innovating Alzheimer’s Drug Discovery Using AI and Knowledge Graphs

Alzheimer’s disease, a formidable challenge in healthcare, has been a persistent adversary, affecting millions worldwide. The complexities of this neurodegenerative condition have introduced significant hurdles in the quest for effective treatments. Traditional drug discovery processes often find themselves struggling with the intricate nature of Alzheimer’s, marked by elusive molecular interactions and multifaceted biological mechanisms.

In light of these challenges, Wisecube, a biomedical AI platform, collaborated with a renowned healthcare institution, St.Johns Cancer Institute at Providence Healthcare, to explore innovative approaches to tackle this pressing issue. This collaboration aimed to leverage the capabilities of artificial intelligence and knowledge graphs, offering a promising avenue to accelerate scientific breakthroughs in Alzheimer’s treatment.

This case study explores how this partnership navigated the complexities of Alzheimer’s, using cutting-edge technology to uncover potential solutions and redefine the landscape of drug discovery for this devastating disease.

Objective: A Dual Approach to Drug Discovery in the Fight Against Alzheimer’s 

The primary objective of this case study was twofold: firstly, to identify new drugs for treating Alzheimer’s disease using Computer-Aided Drug Design (CADD), and secondly, to explore repurposed drugs for Alzheimer’s through the application of artificial intelligence (AI) and knowledge graphs

Lead optimization using CADD

The overarching goal was to employ advanced computational methods to streamline the drug discovery process, enhance lead optimization, and ensure the identified drugs’ efficacy in crossing the blood-brain barrier (BBB). By blending the precision of CADD with the insights derived from AI and knowledge graphs, the collaboration aimed to revolutionize the search for effective treatments, providing hope in the fight against Alzheimer’s disease.

Challenges in Alzheimer’s Drug Discovery: Navigating Complex Terrain

Facing the intricate landscape of Alzheimer’s drug discovery, St. John’s Institute encountered multifaceted challenges in identifying new drugs and repurposing existing ones.  From the complex nature of the disease to the hurdles in finding compounds capable of addressing its complexities, here are some of the obstacles that St. John’s Institute has had to tackle in its quest for effective Alzheimer’s treatment: 

  • Navigating Molecular Complexity: Unraveling the intricate molecular landscape of Alzheimer’s disease proved challenging, demanding sophisticated approaches to comprehend and address complexities in molecular interactions and pathways. Understanding the formation of amyloid plaques by A-beta fragments and the intricacies of tau pathways and their various forms added a layer of complexity.
  • Ensuring Prediction Accuracy: Validating the accuracy and reliability of AI and computational predictions is crucial, requiring strategies to mitigate uncertainties and biases inherent in predictive models.
  • Balancing Speed and Thoroughness: Striking a delicate balance between the swift pace of computational methods and the meticulousness of wet lab validation is essential to efficiently advance drug discovery without compromising reliability.
  • Blood-Brain Barrier Permeation: Overcoming the challenge of facilitating drug permeation through the Blood-Brain Barrier (BBB) added complexity. The extended timeline for developing and approving CNS drugs, taking 20% longer to develop and 38% longer to approve compared to non-CNS drugs, posed an additional hurdle.

Effectively addressing these challenges demanded a holistic approach, integrating AI strengths while acknowledging the imperative for rigorous experimental validation and tackling the distinct challenges associated with CNS drug development.

Wisecube’s Methodology: Three Core Strategies for AI-Driven Drug Repurposing

In the midst of the intricate challenges inherent in Alzheimer’s drug discovery, Wisecube AI proved to be an invaluable ally for St. John’s Institute, playing a pivotal role in overcoming these obstacles. Wisecube’s strategy consisted of three specific approaches:

Wisecube Semantic Discovery Platform

Architecture of Wisecube’s Semantic Discovery Platform (Source: Wisecube)

  1. FDA-approved drug repurposing
  2. Graph-based embeddings and link prediction
  3. Prioritization by blood-brain barrier (BBB) score

This multifaceted process consisted of the following steps:

1. Data Acquisition and Integration

Wisecube collected diverse Alzheimer ‘s-related datasets—genomics, proteomics, and clinical trials, ensuring a comprehensive pool. Integration of FDA-approved drugs, repurposing knowledge graphs, and biomedical literature formed a cohesive knowledge graph.

Data Curation

2. Knowledge Graph Construction

Wisecube consolidated the acquired information into a dynamic knowledge graph at this stage. This innovative system meticulously categorized molecular entities, pathways, and disease mechanisms as individual nodes within the graph. Various methods, including explicit and implicit relationships, were employed for construction of a comprehensive knowledge graph. 

3. AI Algorithm Implementation

Leveraging heuristic and advanced deep learning models, Wisecube deployed state-of-the-art machine learning algorithms. The algorithm, supported by NLP applied to scientific literature, identified contextual relationships and potential drug targets.

Graph Engine

  • Graph-based embeddings: Leveraging knowledge graphs, the team transformed intricate biological data into a string-like format. This facilitated NLP techniques to generate word embedding vectors for each analyzed entity within the graph.
  • Link prediction: One primary approach involved using untyped models, such as Jaccard and DeepWalk. The other employed typed models, like the Trans E-L2 algorithm. Wisecube utilized these techniques to predict connections between knowledge graph nodes, thereby enhancing the graph’s depth and accuracy.

Link Prediction

Deep Walk

  • Assessing BBB permeability with Dryad to rank identified drugs: Post-knowledge graph analysis, a drug list underwent evaluation using an existing BBB model, prioritizing compounds with proven permeability for CNS drug development.

4. Validation, Optimization, and Iterative Refinement

Candidates underwent rigorous silico testing, simulating target protein interactions. The algorithm iteratively optimized lead compounds for safety, efficacy, and blood-brain barrier penetration through meticulous feedback and manual inspection.

Results: Unlocking Advances in Alzheimer’s Drug Discovery

Wisecube’s drug discovery methods, in collaboration with St. John’s Institute, have significantly alleviated the global challenges posed by Alzheimer’s disease. The use of robust AI techniques represents a crucial advance in tackling the complexities of this neurodegenerative condition, with highly significant outcomes.

  • Identification of potential compounds: Advanced methods identified 1000+ compounds targeting the CDK5/P25 pathway, significantly advancing Alzheimer’s drug discovery. Here are sample results displaying identified drugs for biological screening, categorized into different columns. 

Listed identified drugs for biological screening

  • Enhanced understanding of Alzheimer’s pathology: Explicitly focusing on amyloid beta plaques and tau tangles crucial for precise interventions.
  • Contribution to the scientific community: Sharing research findings with St. John’s Institute contributes valuable insights, fostering collaboration and accelerating collective progress against Alzheimer’s.
  • AI-based parallel approaches: This multistep AI-based approach enables de novo research to work concurrently with repurposing, streamlining the drug discovery process without doubling resources.
  • Swift iterations with ranking metrics: Using a ranking-based metric instead of classification allows for rapid iterations based on feedback, expediting the drug discovery process.

With Wisecube’s assistance, St. John’s Institute identified around 8000 repurposing drugs. Ongoing manual inspection is refining the list to narrow down potential drug candidates further.

Conclusion

Through the collaborative efforts of Wisecube and St. John’s Cancer Institute, the project successfully identified drugs inhibiting the clinically crucial CDK5/P25 pathway. Utilizing CADD and AI tools, they pinpointed both NCEs and repurposed drugs disrupting target proteins. St. John’s is now identifying the best leads for further development, enhancing outcomes through advanced drug discovery methods from the thousands of identified compounds. 

This case study shows the transformative impact of AI, computational methods, and knowledge graphs, accelerating the drug discovery process. Furthermore, it reduces costs and improves the potential for successful treatment outcomes for complex diseases like Alzheimer’s. Integrating advanced algorithms and multidisciplinary collaboration with St. John’s Institute sets a new standard, revolutionizing future drug discovery endeavors.

To learn more about how Wisecube streamlined Alzheimer’s drug discovery analysis, head over to our webinar with Dr. Venkata Yenugonda: Curing Alzheimer’s Using AI and Knowledge Graphs.

Table of Contents

Scroll to Top