Taking AI to the next level

by • June 3, 2019 • Feature, Feature-Home, NewsComments Off on Taking AI to the next level148

Canada has become a leader in artificial intelligence research and development, the result of a combination of the efforts of universities, governments, large companies and start-ups. 

Artificial intelligence (AI) is a broad term that refers to computer systems being able to think independently with a human-like intelligence, making decisions based on the massive amounts of data that is fed into the systems.  

This decision-making superpower is being adapted to drive innovation in all parts of the life sciences continuum – from drug design and discovery, through manufacturing and into the clinic. Meet some of the Ontario companies that are using AI to accelerate the pace of discovery and development.  

Using deep learning to accelerate drug discovery 

The lab of the future – that’s how the founders of Phenomic AI describe what they are building with the application of artificial intelligence to interpret reams of biological data faster and more efficiently. 

The company has developed AI-based techniques that quickly interpret large amounts of microscopic data to identify differences between cells and sort them into categories. These types of differences are hard to interpret by eye or with existing computational techniques. 

Phenomic AI’s software aims to understand how thousands of genetic mutations and drugs affect cell health. Scientists can navigate the data interactively and it has already led to a number of important insights in drug discovery programs.  

Traditionally, processing large datasets involves many steps. A researcher would have to segment every single cell in a screen, measure its parameters and use those measurements to produce statistical insights. At a minimum, this this can take two weeks to complete.  

Phenomic AI’s deep learning solutions automate this process. As soon as an experiment is completed, the results are uploaded to the cloud, automatically analyzed and the next day, scientists can explore and interpret the data. “We’re bridging the gap between the time it takes to gather the data and understand what it’s telling us,” says Oren Kraus, co-founder of Phenomic AI. 

The AI ecosystem in Ontario has also provided a particularly fertile environment for the company’s growth. A combination of a large number of qualified scientists, high-quality machine learning programs to train students and investors who understand the AI space have provided a tremendous boost, says Kraus.  

An AI boost for health care decision making 

Each time a patient is admitted to hospital, a clinician must make a multitude of decisions about that individual’s care. Combine that with the continuous changes in drugs, technology and clinical evidence, clinicians face the challenge of keeping on top of all of these changes and using the information to make evidence-based decisions.  

Think Research is helping clinicians make decisions in this complex environment.  

“Care delivery is become more challenging,” says Sachin Aggarwal, Think Research’s CEO. “Think Research is marrying technology and clinical knowledge to help drive the best patient care by maximizing on the decision-making capabilities of clinicians.” 

Think Research’s AI platform works by capturing completed care decisions and linking them to patient outcomes at which point the platform’s algorithms can suggest the best approach to care. Specifically, each time a health care professional makes a decision, it provides a data point that is captured by the platform. This data gets linked to patient outcomes from their medical records. Once the input and outcomes are linked, the algorithms can start to suggest care decisions.  

Think Research is unique because it collects data that is highly detailed about the care decision-making process. Currently, most companies are trying abstract data from the medical records system. “Think Research uses data that comes from real-time decisions,” says Aggarwal. 

It also has access to a high quantity of data, which includes governments and hospitals across the country. This population level data gives the company a huge amount of diversity, variability and volume.  

Think Research’s platform is being used by thousands of clients and facilities and the company has big plans for expansion. “Humans are humans, regardless of where they live in the world and we believe that these decision tools are applicable everywhere in the world. We think we will be one of small number of players in the knowledge-based health care space,” says Aggarwal.  

Polypharmacology: drug discovery for the future 

One protein, one disease, one drug. That is the classical approach to drug design. 

However, the challenge is that drugs do more than one thing when they interact with a cell or are tested in human beings. In addition to targeting the specific protein it is designed to turn off or turn on, a drug can interact with upwards of 300 off-target proteins.  

While this narrow approach gets you to a starting point, it does not adjust for the downstream risk that occurs when you introduce a drug into a complex biological system. In that challenge, Toronto-based Cyclica saw a huge opportunity.  

“The world of polypharmacology – that is all the targets that interact with one molecule – is the world that Cyclica feels passionate about,” says President and CEO Naheed Kurji.  

The company is using a powerful combination of AI-based (knowledge-based) approaches, specifically machine learning and deep learning, to augment computational biophysics (applying computers to model and predict how drugs will interact with the core components of biology) to take a panoramic view of drug interactions with all the pieces of biology in disease development. By helping scientists understand this entire landscape, Cyclica’s approach can help scientists take precise steps in the right direction and decrease drug discovery time, says Kurji. 

Think of Cyclica’s platform like a lock – the biological component for which the drug is being designed, and key – the drug. The knowledge-based approaches use a wealth of data about both the lock and key to form predictions about whether the key will work. Combined with computational biophysics, scientists can gain insight into how the key is going to fit and whether or not it will open the lock.  

This year, Cyclica was named by Investment Ontario as one of the top 10 AI companies to watch and by Deep Knowledge Analytics as one of the top 20 global AI drug development companies.  

Their success is due in part to the research that is taking place at universities, hospitals and institutions like the Vector Institute, and, in the last few years, increasing commercialization of that research, Kurji says. 

“Toronto is now becoming the epicentre for AI in healthcare and drug discovery,” Kurji says.  

Using AI to make AI smarter 

DarwinAI is using artificial intelligence to build artificial intelligence, explains CEO Sheldon Fernandez. 

The Waterloo-based company has developed a platform named Generative Synthesis, the result of years of academic research and was publicly launched in September 2018. The platform uses machine learning to observe a neural network (a type of deep learning), which results in a deep mathematical understanding of the network. The results are used in two unique ways: 

  • They explain how AI software has made a decision, known as explainable deep learning. While AI can perform exceptionally well, it is often difficult to explain why it has arrived at a specific decision.  
  • The results have dramatically reduced the size of the neural network, while maintaining accuracy and reducing inference time (the time it takes to classify, recognize and process new inputs) To date, neural networks have been too large for some applications, such as consumer electronics. This breakthrough brings the potential of machine learning to these types of products. 

“DarwinAI’s founders believe explainability is a critical first step towards ethical AI,” says Fernandez. “We need to understand how and why a machine learning system makes a decision before it can be aligned with a specific moral code.” 

From drug design and discovery to manufacturing and clinical administration, these companies are proving AI is here to stay. 

Comments are closed.