AI in Preclinical Studies: Operational Gains and Scientific Limits

My typical day includes a variety of different things: working with pharmaceutical clients on innovation and strategy, collaborating with global leads to ensure alignment across toxicology disciplines, and liaising with our operations group to define our strategy. At the moment, a big part of these discussions is new approach methodologies and AI.

When I think about AI, I divide it into two buckets:

1. AI for administrative purposes at both the macro and micro levels.
2. AI for the science, including safety assessment.

At last count, we have 15 in vivo sites worldwide. From an administrative perspective, we are evaluating the use AI to schedule and make the best use of our resources across these sites. We are also looking at AI for reporting, including production and interpretation, but this aspect is very much AI-assisted. Subject matter experts still need to review the material and make sure the data is interpreted appropriately.

In terms of science, AI could enable you to put a small molecule structure, target, or receptor into software, and it'll tell you what the on-target and off-target effects are, and things of that nature. Today, that's a dream. It's not close to predicting outcomes with certainty. A cell, or an organism, is a very complex system, and we don't have all the knowledge yet.

We’re not at a place where AI is going to replace animal testing or in vitro testing in the next couple of years.

The administrative and logistical side is where it's having the biggest impact today. In the future, we want to grow that scientific piece.

We see a lot of challenges because “AI” is a scary word. We are starting to use the word automation instead. A lot of companies don’t want us to use AI at all. But we are developing a tool to get you a report within, perhaps, half the time. Wouldn’t you want us to use that? Then, of course, they do want us to do that.

'Sometimes, their contract says no AI, or perhaps refers to no data entry into AI systems. But to write a report, we need data.' — Dr. Steven Bulera.

People get nervous, so we're spending a lot of time working with clients about the language, what we can do, what we can't do, and making sure that the security is there. When we handle data, we aren’t just sending it out to the ethos. We're building tools behind our firewalls, and our information security teams are involved to ensure nothing gets out. 

The data that's being fed into systems is limited, whether it's just that we don't know enough about mechanisms and signal transduction cascades, or whether it’s pharmaceutical companies that are only feeding data into their own systems.

'On that side, nobody is pulling all of the data together. How can they, without losing their competitive advantage?' — Dr. Steven Bulera.

The other limitation is acceptance. Regulatory agencies want to know whether AI was used to generate reports. Some of them really like the idea of AI, but again, it scares others. They wonder, is that expert really reading that document? Are they just dropping it in for the sake of time?

'That's why we talk about AI-assisted interpretations—not just AI interpretations.' — Dr. Steven Bulera.

I can list an entire alphabet of regulatory agencies, each of which may have different opinions. Regulatory agencies also don't talk to each other very well; they may do similar things, but there's no global standard. It's kind of the wild west.

That's why I think contract research organizations are important. We talk to everybody and can be kind of the neutral ground that says, why don't we bring some people together and talk about this? We are a neutral party, and we work with everybody, so we can be an impartial gatherer of information. From our findings, we can go to regulators, saying, ‘Look, we have the support of these global pharmaceutical companies, here's what we're thinking. What do you think?’ And work from their feedback.

At every opportunity, we ask regulatory agencies for their thoughts. Again, this is a new frontier, and the technology is moving faster than regulations. What do you do in the interim? Everyone's trying to figure this out, so it is a challenging time.

Our experience with regulators is that they don't tell you what to do; they ask you to present what you want to do, and they'll tell you if it's good enough or if you need to go back and work on something.

If we can get a number of large pharmaceutical companies to align, it should give regulators confidence.

That was one of my questions for our upcoming summit in October. Do we pull data together? Is there a way to anonymize it? Is there a way to pull data from different companies together so everyone can take advantage?

The other question is, would you be willing to pull all types of data together? Don't give us just the things that have made it to market, we want the failures; that's where the value is. We want everything so we can create a better data set to mine or a tool that's more predictive.

'Generally, scientists all agree on data sharing. Ultimately, it falls apart because of that competitive advantage.' — Dr. Steven Bulera.

If it's all about the patient and meeting an unmet medical need, there should be a little bit of altruism somewhere in there.