Transforming Drug Development Through Real-World Evidence and Artificial Intelligence

Jo Varshney, DVM, PhD, founder and CEO of VeriSIM Life

In this interview, Dr Jo Varshney, founder and CEO of VeriSIM Life, discusses the complexities of drug development, the role of AI in accelerating discovery, the integration of real-world evidence, and strategies to reduce costs while advancing innovation.

Please state your name, title, and any relevant professional experience.

Jo Varshney, DVM, PhD: Hello, I'm Dr Jo Varshney. I am the founder and CEO of VeriSIM Life and our pharmaceutical subsidiary, PulmoSIM Therapeutics. I am a veterinarian with a PhD in comparative oncology/genomics, and I serve on several advisory boards, including the Critical Path Institute.

What are the key stages of drug development, and how do they influence the likelihood of approval?

Dr Varshney: Drug development is very complex, and I don't put this lightly. There are reasons for that because we are dealing with lives. There are many stages, and the stages also depend on the type of drug you're dealing with. Conventionally, and if we are putting this in basic terms, first you have to discover a molecule. Once you discover a molecule, which is also a complex process, you must understand the disease. You have to understand different targets, which need to be specific to the disease so they do not impact healthy cells.

Once you identify a molecule that seems to be promising, you conduct in vitro tests. Essentially, you perform lab tests to determine whether the molecule causes cell death and at what concentration, because if it requires a high concentration, it can create toxicity, and you don't want toxicity. You want the molecule to be both safe and efficacious. Ultimately, those are the 2 main goals for drug development.

If you see positive results, you progress to animal trials, and that's where things get trickier because you have to find the right animal model that accurately reflects the disease and the target of interest.

If that all goes well, you conduct toxicity studies and then submit the data to a regulatory agency, such as the FDA, to seek approval for human testing. This whole process can take more than 16 years and cost over $2 billion. That is the reality we live in.

More than 95% of diseases either lack drugs or do not have drugs that truly work. If you do the math, there are simply not enough drugs reaching the market. The costs are rising, and innovation is suffering with the current fundraising market. There is less investment in innovation and more investment in clinical trials. If we continue this way, there is going to be a huge imbalance where you don't have enough drug assets going to the clinic.

What do you feel are the most significant challenges developers are currently facing during the clinical trial phases?

Dr Varshney: Oh, so many. Finding the right patient group, navigating the recruitment process, the clinical site management process, and getting the right endpoints to ensure that you are not only identifying the drug molecule but also identifying the right patient population, as well as a diverse patient population. Often, that's a big missed point. These complexities really add to cost and time and ultimately lead to significant failure in clinical trials.

How does real-world evidence play a role in the approval process, especially postlaunch?

Dr Varshney: Real-world evidence is very helpful, especially in the time of artificial intelligence (AI), because you can understand how the drug has been effective and responsive in various patient populations.

Often, that's a very big misstep in clinical trials. Why? Because most clinical trials are conducted predominantly in a male Caucasian population of a certain age group to keep it consistent. Postlaunch, you can actually study a lot about how different types of patient populations respond, because not everyone is a 45-year-old Caucasian male. What are the key differences? Ultimately, it helps us understand whether we can make this drug more effective if we adjust the dose, which is a simple thing, or if it's not effective in a certain patient population.

You can learn all these things postlaunch with real-world evidence. It can help inform future clinical trials as well, which is very important. You want the negative data outcomes as well as the positive data outcomes. Positive is easier to get than negative. We need to be very open with real-world evidence to share the negative data more effectively and not look at it as a failed effort because it can really inform so much in new clinical trials or in new drug development processes.

How is AI transforming the early stages of drug discovery, particularly in target identification and lead optimization?

Dr Varshney: I'll start with the Nobel Prize-winning aspects. If the Nobel Prize committee believes that this is a significant achievement that has changed the chemistry of drug discovery, we have to pay attention. Developing more open-source models to help drug developers design better, novel drugs—drugs that might not have been possible otherwise—is key because machine intelligence has a very different way of approaching the drug space than human intelligence. These are ways AI is already making an impact.

We use a hybrid AI approach, which is not as novel in other sectors as it is in ours. We believe in leveraging the knowledge and science that already exist. Like medicine, it has been around for centuries. I believe if we can tap into that knowledge, create simulations using math and physics, and then have AI learn from those, we can create transparency. When an AI makes a prediction, we can ask, "How did you make that prediction?" and it can give you very specific information.

That's very important for human experts, who currently understand biology better than machines, and maybe that will eventually change. For now, if AI makes a prediction that human experts—who understand the disease—can interpret, we can open up a lot more innovation and new drug development approaches to make the whole process more efficient and advance more drugs into clinical trials.

How do you think developers can balance innovation with regulatory requirements for safety and efficacy?

Dr Varshney: I think getting ahead in the conversations is key. Don’t wait until you are in clinical trials. Educating stakeholders and creating transparency around AI models and explainability of AI approaches will help them embrace and advance innovation. However, it’s going to take time, because AI is very broad. There are different flavors of AI.

Certain flavors are easier to digest than others. For example, generative AI presents pretty big concerns because the models are continuing to improve. Where do you stop that improvement? How do you evaluate those continuous improvements? We don't have all the answers yet, but having an open dialogue, transparency, and explainability in AI will help pave the way for greater innovation.

How do you evaluate the cost-effectiveness of integrating technology such as AI into the drug development process? And what metrics or outcomes demonstrate their value to both manufacturers and payers?

Dr Varshney: Time efficiency is the biggest return on investment here. We cannot reduce all the costs because there are certain things you always have to do, but reducing trial and error at the early stage and reducing that cost burden can really improve outcomes, even if we don't change the cost of molecule manufacturing.

Now let's talk about innovation, because today innovation is costly. Innovation means changing current processes and adapting to new ones. In many cases, you either have to hire more people or adopt a technology. I do see that costs are going to be higher initially; however, with more drugs going to clinical trials, the cost of drug development will decrease. I'm optimistic about this. With the right adoption of AI in drug development, we will see a reduction of drug costs, because companies will not be banking on 1 or 2 drugs to succeed. There will be more drugs in the market.

This will create a win-win scenario. It's a win for the companies developing drugs because they will have more drugs in the market, leading to bigger market caps. It’s also a win for the patients, who will benefit from lower drug prices and a wider range of treatments.

I do think that, ultimately, even though there are concerns about AI usage and biases, these concerns will diminish over time, and AI will have a strong positive outcome in the near future. That's what technologies and revolutions have done in the past. There will always be challenges, but overall it will have a huge benefit on drug development, and we're already seeing that. I'm pretty excited to be part of that future.