What the pharmaceutical industry has learned from AI

Listen to the article
9 mins

This audio is automatically generated. feedback.

The pharmaceutical industry has long been eyeing the promise of AI, and predictions for its potential impact are staggering.

AI tools $60 billion to $110 billion Improved operational efficiency, R&D speed and data analytics capabilities are projected to add $2 billion in annual value to the pharmaceutical industry, according to an analysis released in January by the McKinsey Global Institute.

PwC also Potential Benefits of Introducing AI This year’s report provides a company-by-company explanation.

“Overall, pharmaceutical companies that industrialize AI use cases across their organizations have the potential to double their current operating profits by increasing revenue and reducing costs. We expect this industrialization process to start to be fully realized by companies that prioritize AI by 2030,” PwC analysts wrote.

The pharmaceutical industry has historically been slower to adopt AI than other sectors such as retail, but PwC notes that “pharma is rapidly closing the AI ​​value gap.”

As more pharmaceutical companies leverage AI tools, what unique insights will they uncover? Hear from several industry leaders who are using and developing AI tools.

Factors that truly influence test results

Dave Latshaw is CEO of Biophy, an AI company that provides solutions designed to predict clinical trial outcomes and expedite drug development tasks.

“If we could look at everything that influenced the outcome of any trial phase, and there were just a few critical factors that always determined whether the trial was successful or not, those factors would be heavily weighted in the model and the rest would be unimportant. But the truth, and why it’s complicated, is that looking at a huge number of features and the magnitude of their influence on all of them doesn’t reveal one or two absolutely critical factors. They’re all interdependent, and have the same magnitude of influence. This means that a lot of things need to be right for a trial to be successful.

Here are some examples of what may be more important: Choice of endpoints. Often there is a set of endpoints that are typically used, some should be for regulators, but the choice may be a company decision. Some people understand the variability in performance of certain endpoints and are surprised when they differ significantly from expectations. The common wisdom is to use what others use.

But there may be better endpoints that capture the same level of detail and provide the evidence we need, that are more consistent, and that don’t have the risk of inherent mechanisms of variation that can taint the results. For example, in some therapeutic areas, such as psychiatric indications, the placebo effect is often very high, which can lead to an underestimation of the effectiveness of individual drugs. But it is the variability of the placebo, not the drug, that causes the trial to fail. And choosing a placebo exposes you to a degree of randomness that you don’t want.”

Predicting patient outcomes

Mirna Sass is CEO of COTA, a company focused on solutions that leverage real-world data to support oncology research. COTA recently “CAILIN” – a solution suite that utilizes AI It is designed to generate insights from RWD and research datasets.

“You treat a patient, but you want to understand what’s going to happen. Does the treatment stop the disease? Or does it just slow it down? We’re finding we have the ability to predict that, and this is a further analysis than we thought possible. Clinical trial data only tells you a point in time. Here we can get a much more detailed look at the patient’s journey.”

We can now also augment some of our data points with genomics to see which genes are on or off before and after treatment. If we do this over and over again, we can understand patterns in patients. To do this manually would take someone’s entire career, but with algorithms we can figure out which genes are on or off, and, for example, which cancers can be best treated with which biomarker inhibitors. We can almost predict the future.

In the longer term, potential drugs will be taken straight from the discovery stage into real-world models and tested on patient populations to see if and how well they work. The technology is there to make this possible.”

Deeper insight into the timing of gene expression

Craig Thompson is CEO of Cerevance, a biotech company leveraging a large collection of brain tissue samples and AI/ML tools. Developing drug candidates For CNS space.

“What surprised us even more was that we looked at specific genes for each cell type to see if some of them were essentially up- or down-regulated. We always want to use machine learning capabilities to test what we observe.

Some companies look at genes that we’ve identified using machine learning, and then they wonder why they’re going after that target because that gene is temporarily up- or down-regulated at a certain point in the disease progression. In certain cases, our technology has shown us that some pharma companies are going after targets that are not expressed in the human brain, and they’ve made it to phase 2 or 3 trials and failed. On the other hand, in the (neurodegeneration) space, some targets that companies are pursuing are (not understandable) because the genes are very broadly expressed. We expect that there will be a higher incidence of adverse events because of the broad expression, and some of those compounds are in clinical development.”

AI solutions need to be supported by digital infrastructure

Dr. Mert Aral is chief medical officer at Huma Therapeutics, a company developing telemedicine. Patient Monitoring Solutions It also aims to build a platform that can be used for decentralized clinical trials and that any company can use. Build your own healthcare app.

“We’re looking at how we can increase engagement and compliance, and how we can provide patient education and awareness. We know these AI solutions work. We know they can reduce time to treatment change. We know they’re providing data and real-world evidence to pharma companies. But how do we scale?

Today,[solutions like digital health monitoring]still rely on clinical teams. You still have nurses monitoring the data and making calls and engaging with patients. Generative AI allows nurses to monitor the data and trends more efficiently. But we’re thinking: How do we replace those nurses with GenAI nurses so that we don’t have to have as many staff?

Of course, you need to train these models. This is a high-risk environment, and we can’t have a GenAI nurse giving the wrong advice to a patient. We haven’t fully solved that problem yet. But I don’t think that’s the limit from a technology standpoint. We’ll get there, and I don’t think it’s too far in the future.

So when we look at GenAI, we don’t just look at its diverse applications. We also need the right digital infrastructure to launch, train and deploy these solutions at scale, which our health systems are not yet equipped to do.”

AI platforms are unlocking deeper levels of biology and chemistry, creating long-term value.

Ayman Al-Abdullah is a Partner at Mubadala Capital, which manages more than 1,000 ventures. Assets: $24 billionIncludes approximately 45 life science companies.

“The last decade has seen the application of AI to drug discovery and development, with multiple overlapping waves of technology. One view is that the value of AI companies will continue to accumulate in the pipeline they create.

The goal is not to replace scientists, but to empower them with a deeper understanding of biology and chemistry to develop medicines in a more efficient and effective way, while also generating insights that could not be obtained through traditional analytical methods and analysis. The examples are numerous.

One of our investments, Iambic Therapeutics, is using physics-based AI to advance the field of how computational biology and chemistry can be used to discover and develop drugs. Another company, Vevo Therapeutics, is developing large-scale language models for biology to move in vivo drug discovery from downstream stages, limited in terms of representing tumor heterogeneity, into early stage development to discover misunderstood or novel mechanisms of action and novel drug combinations that could potentially be highly effective.”