King Abdulaziz University - University of Oxford Centre for AI in Precision Medicines

On October 26, 2022 I gave a presentation on AI in Healthcare hosted by the King Abdulaziz University - University of Oxford Centre For Artificial Intelligence in Precision Medicines (CAIPM). CAIPM is a leading global innovation center for developing precise therapies using AI and emerging technologies. The Center is a strategic partnership between King Abdulaziz University in The Kingdom of Saudi Arabia and the University of Oxford. CAIPM provides a multidisciplinary hub for building research leadership with a high-impact and sustainable value to provide potential therapeutic solutions to patients around the world. The CAIPM was founded based on Saudi Vision 2030 to improve quality of life which includes promoting health and providing innovative treatment solutions for various intractable diseases for which there are no current treatments.

Highlights from my Presentation

I covered significant milestones that have been achieved using AI across the healthcare spectrum including in radiology, pathology, drug discovery, and gene sequencing. Click here to see highlights for my presentation.

King Abdulaziz University

King Abdulaziz University was founded in 1967 and named for the founder of the Kingdom of Saudi Arabia, King Abdul Aziz bin Abdul Rahman al Saud. KAU enrolls 80,000 students and is ranked first in the Arab world. Educational programs for men and women were established simultaneously at the time of the university’s founding and the university is recognized as a leader in opening the doors of higher education to women in the Kingdom of Saudi Arabia. KAU has an infrastructure of specialized scientific and research entities including supporting deanships, specialized research centers and laboratories, and research excellence chairs. KAU is also interested in protecting, recording, marketing, and transforming promising ideas and inventions into commercial or industrial products.

King Abdulaziz University is ranked first in the Arab world

King Abdulaziz University - University of Oxford Centre for AI in Precision Medicine (CAIPM)

Objectives

• To contribute to the Saudi 2030 vision by achieving a prosperous economy based on digitization?
• To use AI techniques in drug discovery
• To transfer and indigenize technologies to build qualified Saudi cadres to advance scientific research in Saudi Arabia
• To build innovative start-up companies in the field of artificial intelligence
• To strengthen the partnership between research centers and prestigious bodies inside and outside the Kingdom to develop AI technologies and drug discoveries?
• To develop research and patents in the field of artificial intelligence and converting them into products of economic value
• To contribute to enhancing community health

CAIPM Research Areas

1. Charting the disease landscape in Saudi Arabia for drug discovery

The advent of the Saudi Human Genome Project, aimed at sequencing 100,000 individual genomes in KSA, will reveal novel causative genetic hits, underlying both rare and common diseases, which are prevalent in the local population. In this phase, expert clinicians and medical geneticists from KAU and Oxford will work to implement a thorough review of large datasets of genomic and other omics studies, as well as available medical information. The focus will encompass the known biology of the diseases including the genetic transmission, clinical presentation, associated genetic drivers, current treatment options, competitive landscape of novel drugs, and likely financial cost to the healthcare system. Of critical importance, the review will triage a subset of understudied disease genes and targets bespoke to the unmet needs in KSA, for entry into the drug discovery pipeline detailed in phases 2-5.

2. AI-driven novel target discovery

In this phase, experts from CAIPM will apply AI technology to genomic, transcriptomic and proteomic profiles of KSA patients to identify and shortlist novel pharmacological targets, such as those found in Phase 1. This will encompass both artificial neural networks already developed in their laboratory for this task (e.g. Long Short Term Memories, Convolutional Neural Networks), and novel architectures that incorporate newer AI techniques that have been recently proposed (e.g. Transformers, Meta-learning). They will link proposed targets to drugs that exert either intended or unintended effects on the target. Further, we will utilize the Connectivity Map, a resource that has systematically profiled the effect of commonly used medications on each gene, in order to identify drug interactions. They will iteratively feed the validated targets into phase 3, such that these can be further characterized and validated in biological systems. This will enable us to further refine the AI algorithms used in phase 2 by training them on the validation results arising from phase 3.

3. Biochemical characterization for novel targets

In this phase, scientists will assess the biological mechanisms and potential druggability of known, understudied target proteins from the working list (phase 1) and evaluated by machine learning (phase 2). For these targets, enabling tools that include recombinant proteins, structural data, biochemical, biophysical, cellular assays as well as chemical starting points will be generated, to determine molecular mechanism of disease and aid structure-based drug discovery20. The scientists are cognizant that the over-representation of autosomal recessive, thereby largely loss of function, disease-causing variants could present challenges for small molecule therapy. They will devote much effort, using genetic and chemical tools, to establish proof of principle for targets from the pathway of interest that can function as disease suppressors (e.g. substrate reduction for inborn errors of metabolism21), and for targets where rescue of function is amenable (e.g. pharmacological chaperones for misfolding diseases22). These targets will become the focus of our screening campaign via different approaches (enzyme assays, binding assays, crystallography), to generate initial hits for compound design in phase 4.

4. Drug design and discovery for novel targets

Directly following on from phase 3 for targets where there is a potential druggable candidate, computational chemistry and medicinal chemistry, will be conducted in the Target Discovery Institute (TDI) at University of Oxford. The team members have an excellent track record of discovering and validating new modalities of inhibitors for targets with proven genetic implications in the development of disease yet considered to be undruggable. The overarching goal is to design novel small molecules to target dysregulated pathways to be driving the disease phenotype. Early chemical tools will be designed to have high selectivity, potency, efficacy, and cell-target activity to the parameters depicted in the target product profile. Further medicinal chemistry will be performed in the laboratories at KAU to develop one of the chemical probes into a lead with efficacy in a cellular disease model. Compound design will be performed using standard fragment- and structure-based design, structure activity relationship (SAR) analysis and newer methods of AI driven compound design developed in phase 5. New analogues synthesis will be undertaken in the TDI labs and with collaborators at KAU.

5. Development and implementation of drug design AI methodologies

This work package focuses on utilizing cutting-edge AI/machine learning-led drug design approaches in the context of novel disease targets identified in phase 2. Experts from University of Oxford will lead this work package combining their expertise in machine learning, structural bioinformatics, and AI to these datasets. The scientists will build on recent advances in structural cheminformatics which have demonstrated the ability of deep learning AI to use contextual ligand information, to effectively design novel molecules that enhance fragment potency. Small molecule/fragment screening outputs from phase 3 and chemistry insights from ongoing phase 4 will form the main sources of rich data for the basis of methods developed and used in this phase. Prioritised molecule designs from phase 5 will be fed back into phase 4 for synthesis/purchase and phase 3 for testing in biochemical/cellular assays. These results will be fed back iteratively into phase 5’s models to further refine and enhance our understanding of target-specific opportunities for optimising selectivity and potency as part of chemical probe discovery.

Subscribe and Comment

I'm interested in your feedback - please leave your comments. To subscribe please click subscribe at the top of this article.

This article was written by?Margaretta Colangelo. ?Margaretta is a leading AI analyst based in San Francisco. She serves on the advisory board of the AI Precision Health Institute at the University of Hawai?i?Cancer Center.

Twitter?@realmargaretta