When it comes to tech transformations, the pharma industry is often a few steps behind other sectors. However, according to venture capital and investment firm Andreesen Horowitz (a16z), the shift to artificial intelligence in life sciences is not long in coming.
The sector is in the “early innings” of a significant change in R&D driven by AI, a16z noted this week in a blog post co-authored by the head of its bio fund, Vijay Pande.
Consider the escalating pace of AI research and AI-driven pharma partnerships. During several of the company presentations at this week’s J.P. Morgan Healthcare Conference in San Francisco, biopharma CEOs touted generative AI delivering efficiency gains in drug development.
All of these are “leading indicators” of what the VC firm called “an imminent and substantial transformation in the life sciences ecosystem.”
AI promises a future in which computers design the majority of new drugs. With ChatGPT and other large language models (LLMs) increasingly being brought to bear on health and biotech, that vision may be close at hand.
As any venture capitalist would say, though, the future we want must be willed into being. A good initial step involves figuring out on which tasks algorithmic technology can have the biggest impact.
Borrowing “jobs theory” – a framework created by Harvard Business School’s Clay Christensen for determining customer needs – the a16z authors opined on the best “jobs” pharma scientists could “hire” AI to do.
1. Chief among them is target selection.
Pharma is a high-failure rate industry. Discovering, developing and gaining regulatory clearance for a marketable molecule can take about a decade and cost more than $1 billion.
AI can help reduce that rate by helping scientists know which pipeline assets to prioritize, and which not to. That could help speed the pace of bringing high-quality medicines to the clinic.
2. Other labor-intensive, high-cost processes that lend themselves to AI, according to the authors, include human pathway biology.
This involves sifting through volumes of research literature to identify promising avenues for investigation and generate hypotheses for scientists to pursue.
3. Preclinical development, in which animal studies are used to pinpoint compounds and suss out toxicities, holds much potential, as well.
The a16z team calls for tech that can improve the odds of making it across the so-called “valley of [developmental] death,” or the gap between early-stage biomedical assets and later-stage, de-risked projects that can attract biopharma companies and investors.
Virtual models, for instance, can cut down on the need for real-world animal studies by showing which agents will translate to clinical compounds. Algorithms can replace the bespoke, trial-and-error approach. The era of “serendipitous biologics discovery” can become a thing of the past.
4. Moving to the clinical trial phase, which is responsible for half the total investment cost of drug development, AI can save yet more time and money. Think about the enormous amount of data generated by these studies.
“Even modest improvements in the way these data are analyzed could slash costs and timelines for drug development,” per a16z’s blog post. Opportunities for pharma range from LLM-guided clinical trial design and protocol drafting to AI-driven patient selection, data analysis and regulatory submission.
5. Broader areas in which AI plays a role include optimizing manufacturing processes, helping the industry hit its ESG goals and inventing brain-computer interfaces that may help restore lost neural functions, the VC firm pointed out.
The last decade has witnessed an unprecedented rise in computing power and with it, improvements in many aspects of drug development. Of late, advancements in natural language processing, like genAI, are being applied to biopharma.
Tech can’t replace science. But to the extent it becomes more infused in the aforementioned jobs, it can accelerate the work scientists do every day to discover and develop biomedical breakthroughs for patients.
