Artificial intelligence (AI) far too often pops up as a term used vaguely to refer to any process that appears to involve more computers than it did twenty years ago. But concrete examples of how this informatics technology can improve fields like life sciences are harder to come by. I recently spoke to Krishnan Nandabalan, founder, CEO and president of InveniAI, which aims to use AI techniques to more quickly identify pharmacological compounds and get drugs to patients faster, to get the full picture.
Ruairi Mackenzie (RM): How would you like to set the record straight about AI in the life sciences?
Krishnan Nandabalan (KN): AI is used as a buzzword now. Everyone is talking about AI and machine learning. Sometimes they use it interchangeably or they’re using it whilst not knowing exactly what they mean. That means there is a lot of confusion over how AI is currently used in our lives. When you type in something into Google or when you go to Amazon, you don’t realize that there’s actually AI involved in that. AI is embedded now into a lot of the automation that is currently used in life sciences research. When companies start using AI-based processes, that’s good. I think it can bring in efficiencies, but AI in that sense is not going to solve the bottlenecks that currently exist in drug discovery and in the life sciences.
But when you think of using AI and machine learning to solve these problems, for example methods to use all the available data that was not available five years ago: real world data from patients, expanded clinical trial data and ever-expanding scientific publications. The real promise of AI and machine learning is that we can use all of this data comprehensively, as well as in ways that were not considered probable even a couple of years ago.
There is certainly a lot of hype about AI. The impression that the hype leaves you with is that everything is state-of-the-art right now in terms of application. That is not true. If your aim is to deliver to the consumer what already exists, then yes, the state-of-the-art exists, but if it is to deliver the promise of AI and machine learning in developing new drugs or devices or patient solutions, then that’s a constant process and people should accept the fact that it may take time to see the results.
An analogous example is the Human Genome Project. The first human genome was completely sequenced in the early 2000s and people thought that we were going to see a flood of new drugs. And when that did not happen in the next two, three years, they were disappointed. But if someone took the effort to understand the origins of all the new drugs and therapies in cancer, their origin lies in the genome project, we will see the benefits, but it’s going to take time.
RM: What are the steps required to take us from now to the promise of the future?
KN: At InveniAI, we are already using artificial intelligence and machine learning to re-innovate drugs that are either old drugs not being sufficiently used or drugs that are stalled in phase 2 or beyond phase 2. We know how the drug works and that it is safe in human beings, but it just didn’t go all the way through and get approved. The usual reason for that is incomplete science, incomplete understanding of the disease pathology. That is changing on a daily basis. AI is a very efficient way of monitoring these changes and then understanding those changes to see if we can use existing solutions to meet these unmet needs.
That’s one thing the industry is currently using AI for. Another is in diagnosis. There is a very good example in how institutions like Memorial Sloan Kettering and the MD Anderson Cancer Center are using AI to standardize diagnosis of cancer so that other centers that don’t have their expertise can follow the same standards of diagnosis and treatment. It’s a great thing.
In manufacturing, especially of complex products like biologics and antibodies, AI is very efficiently used in monitoring systems to see if anything can go wrong, is going wrong, is about to go wrong; you can save a lot of money and precious time. Because patients are waiting for these drugs, that’s very important.
Finally, there’s literally petabytes of clinical trial data sitting with big pharmaceutical companies and I think now we have the tools based on artificial intelligence to mine that data so that we can run efficient clinical trials. These trials won’t take as long or won’t need as many patients as before, so that they can more quickly go to completion and approval.
I think all of these aspects that I’ve illustrated have already happened and of course these are the first iterations. I would imagine that future iterations would be better, faster, cheaper and so on. I think where the really exciting piece of work that’s happening is in designing new drugs, which used to be the domain of computational and structural biology. Now we are applying artificial intelligence and machinery to those fields and coming up with novel drug designs. This is very promising and very exciting but it’s going to take time to see the results of such efforts because while you can design new drugs you still have to go through the clinical development process. That doesn’t stop.
RM: Do you have any examples of drugs that have been identified through AI analysis of clinical trial data that have been repurposed in this way?
KN: An example is thalidomide, which Celgene repurposed in cancer. This is a drug that was taken off the shelf because it was causing birth defects. It is possible to repurpose drugs, but the question remains, “How do you industrialize it?” and that’s where AI comes in. We at InveniAI have identified two drugs. One is an approved pre-anesthetic sedative called Precedex which has an active ingredient called Dexmedetomidine. Precedex is currently used as an IV agent and as a pre-anesthetic sedative in the ICU.
We were looking for agents that would cure hyperarousal, because hyperarousal causes not just insomnia but aggression and agitation in many neurodegenerative diseases including Alzheimer’s disease, schizophrenia and PTSD.
By using AI we were able to link up pathways, pathophysiology and this drug mechanism and we realized that by creating a different formulation of the drug, a sublingual ten-fold delivery, we could use it for treating acute agitation. We tested it out first in a pre-clinical model and then clinical trials in phase 1b, and phase 3 is starting this month. We hope to have the results in the middle of 2020. Here’s a perfect example where just by using available clinical trial data, pharmacological data, biological data we were able to reposition an existing drug with a completely new application.
The point I’m trying to make is that this is no longer theory, this is entering late-stage clinical development and perhaps if these drugs are all doing the right thing, by next year you’ll have the first drug approved that was picked by our AI system for reposition.
Krishnan Nandabalan was speaking to Ruairi J Mackenzie, Science Writer for Technology Networks