Leveraging the impact of diversity in neurodevelopmental disability by integrating machine learning in personalized interventions.

Neurodevelopmental disability (NDD), which is an umbrella term for autism, attention deficit, and intellectual and learning disability, affects 13% of the population. It has major economic and quality-of-life impacts on NDD individuals and families, and substantial economic burden on the healthcare system. So far, treatment is aimed only at general symptoms, which often leads to low efficacy and frequent side effects.

The advent of novel genetic testing methods has provided plenty of evidence of the major impact that genes and their regulation have on clinical presentation in NDD. Nonetheless, there is a large diversity among individuals with NDD, even with the same genetic mutation. This is not unique to NDD as it is seen widely in many other medical conditions. The complexity derived from the genetic heterogeneity and the clinical (neuro) diversity has proven challenging to traditional approaches for treatment.

Recent research in the UK and Canada has led to the development of large databases recording detailed information about individuals with NDD. Artificial intelligence (AI) now provides us with the tools to quickly analyze the information in those datasets. In particular, we will use machine learning (ML) to manage complex information, leading to the acceleration and better prioritization of interventions. Also, our project takes a novel view on the understanding of genomic information in NDD. Instead of directing our focus only on exploring data from a single individual or small group of individuals carrying the same gene mutation, our team will apply ML to large databases to identify features (from genes and their biology) correlated with improved clinical outcomes.

In addition, we will use ML to better understand the interdependence between different symptoms to develop treatments that have a globally positive impact. In other words, we would find solutions that improve cognitive skills without impacting sleep negatively or generating more anxiety, as has been seen in previous clinical trials.

We will finish by providing the entire scientific community with an open access portal, including our research findings, which will be integrated with the current Open Targets platform, a partnership between academia and industry in the UK that allows researchers to access linked data on diseases, genes and drugs in a single site. Researchers will be able to provide further information, which will improve the ML model.

To ensure that we accomplish our objectives, we have assembled a team of experts in clinical and genetics of NDD: Dr. Bolduc (Canada); in computer science of genomics, molecular and pharmacological data: Dr. Dunham (UK); bioinformatics: Dr. Droit; machine learning: Dr. Greiner; social sciences, patient engagement and health economic: Dr. Zwicker. Our team has also developed strong links with NDD patient and research organizations in Canada and the UK, which will provide insight throughout the project. We are supported by collaborators involved in family and government engagement, ethics and data management in the UK and Canada. The project will also be a unique opportunity for multidisciplinary international training.

Our project will show how ML can disassemble the complexity and diversity seen in NDD to develop more successful interventions. It will allow us to develop new ML approaches that will be readily applicable to other disorders where personalized interventions have been lagging behind diagnosis. More importantly, it will bring together families, society and scientists into a shared space where more and better information is exchanged. Finally, our project will embrace responsible implementation of data privacy and confidentiality while recognizing the need for data sharing to develop better interventions.

Our project will increase awareness of the positive impact of machine learning (ML) in developing treatment informed by patients. While most people associate ML with self-driving cars and facial recognition, its enormous impact in pharma remains largely unknown to the public. Yet ML can help us to quickly process and reliably exchange vast amounts of accurate, relevant and timely information amongst an array of diverse knowledge users. Moreover, the techniques applied here will be immediately transferable to research on the genetic and phenotypic influences on other rare and common disorders.

Our project will show that ML can enhance expert ability to understand complexity and diversity related to neurodegenerative disability (NDD). Affecting 13% of the population, NDD represents a large group of genetically heterogenous disorders with overlapping clinical features 1,4. Thus, the development of drugs for NDD is extremely slow and costly. Our project will show how ML can process large datasets and identify the important targets for treatment in a maximum number of individuals.

Our program will showcase that ML allows for a rational and cost-effective prioritization of candidate treatments. The main scientific impact of our project will be to allow researchers to prioritize candidate treatments for NDD using the input of human genomic data, and limit the unnecessary exposure of children with NDD to drugs. In addition, it will avoid the repeated failure of clinical trials that rely too much on trial and error.

Our project will illustrate how confidentiality and privacy are respected while using ML responsibly. With recent events, including the use of large amounts of data and AI, a negative view of ML has developed with the public about respect of privacy. Our team therefore includes a major focus on ethics and individual privacy. It's a focus that we share with all investigators and knowledge users because we believe that, not only must privacy be ensured, but it must be seen to be ensured. We will also develop protocols, in collaboration with data privacy expert Dr. Mouratidis (Brighton,UK) for harmonization between UK (General Data Protection Regulation-GDPR)4 and Canadian (Personal Information Protection and Electronic Documents Act-PIPEDA) 5 datasets that will serve as a model for future international data sharing.

Our project will build capacity in our understanding of the genomic basis of NDD. Dr. Dunham, as Director of Open Targets has created a unique platform combining the functional data necessary for this project. In his clinical practice and research lab, Dr. Bolduc has developed a very successful rapport with NDD individuals and their families, especially understanding their needs and conditions. His lab has also developed good relationships with artificial intelligence (AI) experts in searching for data on NDD. By connecting with international colleagues and using machine learning on large databases, Drs. Dunham and Bolduc and other team members will substantially accelerate their capacity building. The result, in both the UK and Canada, will be the training of several highly qualified people, development of better investigative techniques and significant advances in our understanding of NDD.

Finally, our project will build synergies between researchers/clinicians and families of those with NDD and allow for a sustained development of data stored and managed responsibly.