AI and the End of Clinical Trials

AI and the End of Clinical Trials

The future of medicine may be in the computer lab.

View this email in your browser

AI and the end of the clinical trial?
With China’s struggle to contain the coronavirus, or the Novel Coronavirus Pneumonia (NCP), at the forefront of many peoples’ minds, many have become increasingly curious about how new drugs and vaccines are developed. This week, we’re learning about how AI could potentially improve it.

We’ll also pay homage to The Sims franchise, provide an update on the California Consumer Privacy Act (CCPA) which is now in effect (for a primer on it, check out our overview here), and investigate the return of…the catalog!

We hope you’ve been enjoying this newsletter and would love any feedback (erica@nycmedialab.org). Thank you again for reading!

Best,
Erica Matsumoto
NYC Media Lab

In recent weeks, the novel coronavirus that originated in Wuhan, China at the end of 2019, now renamed “severe acute respiratory syndrome coronavirus 2” or “SARS-CoV-2,” has brought drug development into sharp focus for many. From the first days of the outbreak, multiple groups have been working feverishly to develop a vaccine or cure for the novel coronavirus. In some cases, research teams are using AI in their drug discovery efforts.

While the use of AI to develop coronavirus therapies is top of mind at the moment, the use of AI in pharmaceutical development is a burgeoning topic of interest across all therapeutic areas. In fact, the first-ever entirely AI-developed drug, DSP-1181 (a long-acting and potent serotonin 5-HT1A receptor agonist that’s being investigated as a candidate for treating obsessive-compulsive disorder), is about to begin clinical trials in Japan. This drug was created through joint research between Japanese pharmaceutical company Sumitomo Dainippon Pharma and Exscientia, which provided the discovery platform, Centaur Chemist AI.

Exscientia is incredibly bullish about AI’s future impact on drug development. Its CEO, Andrew Hopkins, says, “We believe that this entry of DSP-1181, created using AI, into clinical studies is a key milestone in drug discovery.” The company’s Twitter account is even more bombastic, tweeting:

Source: Exscientia Twitter

These pronouncements, as well as the use of AI in fighting NCP, are exciting. And certainly, it seems that some pharmaceutical companies share Exscientia’s belief. Both Sanofi and Bayer have entered multimillion drug development deals with the company. But how exactly is AI being used to develop new drug therapies?

HIGHER-POTENTIAL INVESTIGATIONAL COMPOUNDS MORE QUICKLY

First, it’s important to understand how pharma businesses are using AI in drug development. Companies are using AI to achieve to:

• Predict or identify untested components or compounds that researchers should explore to find new cures
• Predict new treatments’ potential effects
• Support clinical decision making
• Understand how drugs interact with the body’s proteins and therefore determine and develop polypharmacological profiles more quickly and at a lower cost

Novartis is a leader in the use of AI for drug discovery. It’s currently using machine learning to classify digital images of cells treated with different experimental compounds. The hope is that by collecting and grouping compounds with similar effects together before passing the clean data to researchers, these algorithms can produce insights that researchers can leverage in their work and reduce time spent testing ultimately ineffectual compounds (at present, testing compounds against samples of diseased cells accounts for a significant amount of time and money spent in drug discovery).

For an in-depth look at how AstraZeneca is using AI for drug discovery, check out a talk some members of its AI development team gave at the 2019 GAIA Conference:

IMPROVED CLINICAL TRIAL SUCCESS AND REDUCED COSTS By far the most important potential benefit that AI offers to pharmaceutical companies is improved clinical trial success rates. According to a 2018 study published in Biostatistics, only 13.8% of all drug development programs eventually lead to an approved product. Thanks in part to this high failure rate, the Tufts Center for the Study of Drug Development (CSDD) estimated that the average cost to bring a new drug to market was just under $2.6 billion in 2017.

In certain disease areas, the cost of new drug development is even higher than the general failure rate: for new drugs targeting complex and/or poorly-understood conditions such as Alzheimer’s disease, failure is nearly guaranteed (from 2002–2012, the failure rate for such drugs was 99.6%).

It doesn’t take a genius to figure out that reducing the failure rate of new drug development programs could go a long way toward reducing the cost of developing new drug therapies in general. Admittedly, there are other causes of rising drug costs, including lack of generic competition (a perennial political issue in the U.S.). However, it’s undeniably true that improving returns to investment in drug discovery (which were a measly 1.5% in 2018, compared to a 10.5% cost-of-capital) could help make new drug development sustainable for pharmaceutical companies. In fact, Novartis CEO Vas Narasimhan estimates that technology could potentially reduce clinical trial costs by up to 20%.

In addition to reducing costs, AI can also help generate hypotheses that might not otherwise occur to scientists. Because traditional drug discovery is dependent on scientists’ judgments and constrained by practical reality (only a certain number of experiments can be feasibly funded and run in parallel), there’s a heightened risk of bias and blind spots. By drawing on large data sets to generate insights, AI methodologies can help researchers avoid the implicit bias that arises when only limited, local data is used to draw a conclusion, thereby revealing new avenues for exploration. A WORD OF CAUTION However, despite the exciting developments in AI’s use in pharma development, it’s still important to take pronouncements that AI is the future of pharma development with a giant grain of salt. Narasimhan, for one, warns that there’s still “a lot of talk” but “very little in terms of actual delivery of impact” in terms of AI and ML use in drug development programs.

Narasimhan also expresses skepticism about the idea of using real world data to replace traditional randomized controlled trials. In a 2019 Forbes interview, he said:

“I do believe that the power of randomization, the power of blindedness, it’s what enables us to control for all the things we don’t know about [the]complexity of human life and human biology. To think we’re going to take that away, and then be able to really determine the efficacy of a medicine, puts a lot on the statistics that I don’t think we have. I’m more of a real world evidence realist, after we have randomized placebo controlled data that really tells us that something has the effect we think it does, then to explore more effects or explore more uses through real world evidence makes a lot sense, but I don’t see this as a panacea that suddenly will make the world much easier.”

There’s also a risk that AI can be wrong. Alix Lacoste, Ph.D., vice president of data science at BenevolentAI (which is one of the two groups currently using AI to find compounds to fight SARS-CoV-2), explains, “Sometimes the algorithm gets it wrong. In target identification, for example, the AI algorithm may have issues distinguishing between potential positive and negative biological effects on the disease course, or predicts drug targets that scientists know will likely have significant side effects.” In such cases, there’s no substitute for human intervention to tell the AI to filter out specific drug or target classes. In this sense, AI in the life sciences is the same as AI in any other field — it must be trained diligently to ensure that it returns useful results.

Once fully realized, the use of AI in the pharma industry is likely to help speed up pharma development, but is unlikely to ever fully replace human scientists. Lacoste envisions a beneficial partnership between the two: “The role of artificial intelligence, whether it is applied to identifying targets, designing new drugs, or repurposing old ones, is to augment scientists’ abilities, not replace them. Scientists play essential roles in determining the data to use in machine learning and in providing expert evaluation of the results, both for additional accuracy and nuance.”
How The Sims navigated 20 years of change to become one of the most successful franchises ever When most people think of successful franchises, it’s unlikely that The Sims will come to mind. However, through its four mainline games and expansions, the Sims franchise recently reached 200 million copies sold on PC. Globally, there are 20 million unique The Sims 4 players worldwide. The Sims’ success is owed in large part to the Sims team’s ambitious vision for the game’s universe and constant efforts to ensure that the Simsverse reflects the world that players themselves live in. By making the The Sims as lifelike as possible, the Sims team has created a franchise that stands the test of time and continues offering players a reason to return. 13 min read California’s Privacy Law Is Finally Here. Now What? The California Consumer Privacy Act (CCPA) went into effect on January 1, 2020, giving California residents powerful new privacy protections, some of which could be extended to consumers across the U.S. Consumer Reports offers a useful guide to what the CCPA means for individuals and companies. 10 min read Why Catalogs Are Making a Comeback

In a counterpoint to those who believe digital marketing is the only way to reach consumers where they are today, catalog mailings have been steadily increasing since 2015. At the same time, response rates have improved, increasing by 170% from 2004 to 2008. Consequently, numerous brand and retailers, including Nordstrom, Patagonia, Crate and Barrel, Restoration Hardware and even pure-online retailers such as Wayfair, Bonobos, Birchbox and Amazon are printing catalogs.

What explains this phenomenon? Research by Colorado State University marketing professor Jonathan Zhang suggests that catalogs are an important expression of a firm’s creative and aesthetic capabilities, as well as its ability to emphasize with consumers and evote emotional connections that drive purchase.

What we want to know is this: does this mean Delia’s — arguably the pioneering force in catalogs’ original rise as a teen marketing tool in the 1990s — could make a triumphant return to our lives?

7 min read This Week in Business History

February 13, 1959: Barbie makes her debut at the American Toy Fair in New York City

By the end of the 20th century, Mattel’s flagship doll will become a $1.9 billion industry. Ruth Handler, Mattel’s cofounder, designed Barbie after observing that her daughter Barbara preferred playing with the more shapely paper dolls in her collection.

Even now, more than 60 years after her debut, Barbie continues to make news. At the end of January, Mattel unveiled new Barbies featuring vitiligo and hair loss to better represent “a multi-dimensional view of beauty and fashion.”

Source: Paul Jordan/Mattel For more about Barbie’s history, check out “Tiny Shoulders: Rethinking Barbie,” which explores the doll’s invention, multiple reinventions and controversies over the years.

This email was sent to <<Email Address>>
why did I get this? unsubscribe from this list update subscription preferences
NYC Media Lab · 370 Jay Street, 3rd floor · Brooklyn, New York 11201 · USA