"Artificial Intelligence for Design (AI for Design) is one in a series of 4 interlinked National Research Council of Canada (NRC) Challenge programs that address research objectives of national and international importance," says Kevin Thomson, Program Director, AI for Design Challenge program. Each program features master projects; the program includes gene and cell therapies, communications and clean fuels among others.
"In collaboration with leading academic and small and medium‑sized organizations, we are challenging ourselves to develop AI tools that vastly accelerate and advance Canada's capacity to design scientific and engineering innovations." AI algorithms and datasets allow faster and more intelligent search of design spaces, leading to massive economies of scale and cost reductions. They also improve the relationship between computer simulations and physical experiments—and speed up discoveries in the master projects.
"Such collaborations strengthen the technology ecosystem for the benefit of all Canadians," adds Thomson. "We encourage expressions of interest from qualified collaborators or stakeholders whose vision of the future aligns with ours." These include anyone from Canada's most seasoned researchers to postgraduate students.
Core AI for Design master project
This project performs exploratory research in machine learning (ML) and AI to develop new methods for scientific design. It serves the needs of the other AI for Design projects to speed the application of AI.
"AI is particularly important in speeding up the drug‑discovery process, for example," says Harry Guo, Research Officer at the NRC's Digital Technologies Research Centre. "Since most of that data is structured on graph neural networks (GNNs), using AI graph illustrations can improve our ability to quickly predict which molecular properties among billions can be used to develop solutions for specific challenges."
Traditionally a long, expensive activity, drug development includes complex research and testing, as well as lengthy approval procedures by authorities such as Health Canada and the U.S. Food and Drug Administration. According to Jian Tang, Assistant Professor at Mila and HEC Montréal, developing a new drug takes more than 10 years on average and costs about US$2.5 billion. "AI is a huge opportunity to accelerate the process," he says, estimating that within the next 5 years, harnessing the power of AI could reduce the time from 10 years to 1.
"GNNs can also probe uncharted territory, opening doors to new material discovery, advanced circuit design and novel drug invention among others," Dr. Guo adds. By developing new graph designs, structures and properties, and using AI and ML to mine information from large data sets, researchers have uncovered immense potential for solving real‑world problems. And in the race to find a COVID-19 vaccine, this has enabled them to discover new molecules for drug development.
AI-assisted photonics design master project
Technologies such as the Internet of Things (IoT), 5G and others demand enormous increases in connectivity and bandwidth. These call for unprecedented innovations in nanophotonic component design. "As the complexity of photonic devices grows, the traditional approach to design becomes impractical," says Yuri Grinberg, an NRC research officer and lead for the AI-Assisted photonics design master project. "Such global challenges require a major shift into leveraging powerful AI tools to master the complexity of next-generation devices."
Innovative research by the NRC aims to equip nanophotonic experts with the ability to deal with high-dimensional design problems and competing performance objectives. AI tools will not only simplify their tasks, but also provide unprecedented insight into the physics of light at the nanoscale.
Grinberg points out that success in this area takes strong multidisciplinary collaborations such as the recently established partnership between the Advanced Electronics and Photonics and Digital Technologies research centres through the High-throughput and Secure Networks and Artificial Intelligence for Design Challenge programs. The AI-Assisted photonics design master project supports the High-throughput and Secure Networks Challenge program's goal to develop innovative technologies that enable service providers to offer extreme yet affordable broadband services in rural and remote communities in Canada.
A first success in this line of work was featured in the Nature Communications magazine article, "Mapping the global design space of nanophotonic components using machine learning pattern recognition." The authors demonstrate a machine learning-based approach to efficiently map and characterize the complex, high-dimensional design space of a nanophotonic interface between an integrated optical circuit and optical fibre.
Biological systems master project
"In gene and cell therapy research, AI is the great enabler," says Dr. Miroslava Cuperlovic-Culf, Senior Research Officer and Team Lead at the NRC.
"AI-enabled research will have an unprecedented effect on health care, since it not only makes research more efficient, but also allows us to solve problems that we would otherwise have no hope of tackling." For example, powerful immunotherapy for cancer is a complicated, time-consuming process that costs about $500,000 per treatment. Using AI, researchers can cut those costs and time by designing therapies through deeper analysis and simulation of behaviour of therapeutic cells in different environments—allowing their development and application in a safe and predictable way. The team also aims to apply AI for design of therapy delivery methods that are essential for treating currently incurable neurological and neurodegenerative diseases.
Collaborators in this master project include Dr. Steffany Bennett, University of Ottawa Research Chair in Neurolipidomics and Dr. Irina Alecu, a postdoctoral research scientist in Dr. Bennett's lab at the University of Ottawa who leads the Discovery and Target Research and Development Programs. Bennett, who researches new Alzheimer and Parkinson's therapies by modulating lipid (fat) compositions, points out that the NRC's ability to manage the immense amounts of data gathered from thousands of patients, and create tools that can be used immediately, accelerates research as never before. "What we have done in this collaboration over the past 6 months would normally take about 4 years."
Cuperlovic-Culf adds that while the NRC specializes in applied research, their ability to work with interdisciplinary basic research partners ensures that all relevant skills in any project are covered. Their work so far on this project has already generated immense interest from clinicians who can see the benefits of their results as well as the AI tools the teams are developing.
Deep materials science master project
This project supports the goal of the NRC's Materials for Clean Fuels Challenge program to make renewable fuels out of materials such as clean electricity, water and carbon dioxide (CO2). The work will facilitate the decarbonization of Canada's oil and gas and petrochemical sectors.
This includes developing AI tools that enable scientists to better predict material properties and intelligently search vast material spaces to identify—or invent—catalyst materials for hydrogen production and carbon dioxide conversion into synthetic fuels.
"Deep learning accelerates this materials discovery process using robotics, AI and high-throughput experimentation," says Isaac Tamblyn, an NRC research officer and lead for the deep materials science master project. For example, current methods for identifying new molecules or materials consume millions of hours of computer time. AI can reduce those hours to seconds by using neural networks. It can also suggest new materials based on blends of existing properties, and be applied in any field to help manufacturers advance product design.
"Canada has unique talent in academia, industry and government for fundamental AI research that puts us at the leading edge of developing new AI methods," says Tamblyn. "And we're using that advantage to apply them to real-world engineering problems."
This article in Science Direct magazine describes how mapping materials science problems onto computational frameworks suitable for machine learning can accelerate materials discovery.