Kidney Week

Abstract: PO0657

Artificial Intelligence-Driven Target Identification in CKD

Session Information

• CKD Mechanisms - 2
  October 22, 2020 | Location: On-Demand
  Abstract Time: 10:00 AM - 12:00 PM

Category: CKD (Non-Dialysis)

• 2103 CKD (Non-Dialysis): Mechanisms

Authors

• Kumar, Chanchal, Translational Science and Experimental Medicine, Research and Early Development, Cardiovascular, Renal and Metabolism (CVRM), BioPharmaceuticals R&D, AstraZeneca, Gothenburg, Sweden

Chanchal Kumar,

• Buvall, Lisa, Bioscience Renal, Research and Early Development, Cardiovascular, Renal and Metabolism (CVRM), BioPharmaceuticals R&D, AstraZeneca, Gothenburg, Sweden

Lisa Buvall,

• Ryaboshapkina, Maria, Translational Science and Experimental Medicine, Research and Early Development, Cardiovascular, Renal and Metabolism (CVRM), BioPharmaceuticals R&D, AstraZeneca, Gothenburg, Sweden

Maria Ryaboshapkina,

• Woollard, Kevin, Bioscience Renal, Research and Early Development, Cardiovascular, Renal and Metabolism (CVRM), BioPharmaceuticals R&D, AstraZeneca, Cambridge, United Kingdom

Kevin Woollard,

• Williams, Julie, Bioscience Renal, Research and Early Development, Cardiovascular, Renal and Metabolism (CVRM), BioPharmaceuticals R&D, AstraZeneca, Gothenburg, Sweden

Julie Williams,

• Kurbatova, Natalie, Quantitative Biology, Research and Early Development, Discovery Sciences, BioPharmaceuticals R&D, AstraZeneca, Cambridge, United Kingdom

Natalie Kurbatova,

• Reznichenko, Anna, Translational Science and Experimental Medicine, Research and Early Development, Cardiovascular, Renal and Metabolism (CVRM), BioPharmaceuticals R&D, AstraZeneca, Gothenburg, Sweden

Anna Reznichenko,

• MacPhee, Iain, Early Clinical Development, Research and Early Development, Cardiovascular, Renal and Metabolism (CVRM), BioPharmaceuticals R&D, AstraZeneca, Cambridge, United Kingdom

Iain MacPhee,

• Bohlooly, Mohammad, Translational Genomics, Research and Early Development, Discovery Biology, BioPharmaceuticals R&D, AstraZeneca, Gothenburg, Sweden

Mohammad Bohlooly,

• Paul, Dirk S., Centre for Genomic Research, Research and Early Development, Discovery Sciences, BioPharmaceuticals R&D, AstraZeneca, Cambridge, United Kingdom

Dirk S. Paul,

• Dunbar, James, BenevolentAI, London, United Kingdom

James Dunbar,

• Taddei, Andrea, BenevolentAI, London, United Kingdom

Andrea Taddei,

• Davies, Mark, BenevolentAI, London, United Kingdom

Mark Davies,

• McMahon, Siobhan, BenevolentAI, London, United Kingdom

Siobhan McMahon,

• Mori, Andres, BenevolentAI, London, United Kingdom

Andres Mori,

• Fritsche-Danielson, Regina, Research and Early Development, Cardiovascular, Renal and Metabolism (CVRM), BioPharmaceuticals R&D, AstraZeneca, Gothenburg, Sweden

Regina Fritsche-Danielson,

• Bendtsen, Claus, Quantitative Biology, Research and Early Development, Discovery Sciences, BioPharmaceuticals R&D, AstraZeneca, Cambridge, United Kingdom

Claus Bendtsen,

• Laerkegaard Hansen, Pernille B., Bioscience Renal, Research and Early Development, Cardiovascular, Renal and Metabolism (CVRM), BioPharmaceuticals R&D, AstraZeneca, Gothenburg, Sweden

Pernille B. Laerkegaard Hansen,

Background

Involvement of multiple pathways and complex pathophysiology are few of the hallmarks of Chronic Kidney Disease (CKD). These reasons contribute to the challenge for drug discovery in CKD, which is a major contributor to global disease burden. Availability of a wealth of CKD omics data has opened avenues for novel insight generation through unbiased integrative analysis. In a pioneering effort AstraZeneca and BenevolentAI initiated a collaboration to leverage the potential of artificial intelligence (AI) to generate novel hypotheses for drug targets in CKD.

Methods

We have created a CKD knowledge graph (CKD-KG) - a knowledge base of biological and chemical entities (genes, small molecules, etc) and their relationships (gene-disease associations, therapeutic drugs, biological processes, etc) and augmented with CKD specific information derived from both public and AstraZeneca proprietary data sets. The CKD-KG was constructed by aggregating information from structured biomedical databases, machine learning (ML)-based extraction from unstructured sources, and patient-centric omics datasets (i)unstructured: 140M documents, 1B relationships, (ii) structured: 30M relationships, 3B omics data points, (iii) 35 licensed data sources, and (iv) 53 CKD omics datasets. The CKD-KG was used as input to BenevolentAI’s relational inference and causal reasoning ML models to produce target hypotheses for CKD.

Results

The fleet of models identified 295 potential targets that were triaged down to 69 targets. These 69 targets have been further prioritized based on an in-house human target validation pipeline, and additional criteria such as safety and druggability, in line with AstraZeneca’s 5R framework. We are undertaking in vitro studies via genetic modification in selected cell types to generate target-specific CKD-linked readouts. Subsequently, we will employ in vivo studies to confirm the mechanism of action for targets that had shown successful in vitro readouts. Eventually, we will progress targets with compelling novel biology within our renal portfolio.

Conclusion

CKD-KG enables a transformative approach to generating novel target hypotheses with the potential of improving health outcomes for CKD patients.

Funding

• Commercial Support –