Vijay Pande spent more than a decade building one of venture capital’s most prominent life sciences practices. Now he is betting small. After managing close to EUR 3 billion at Andreessen Horowitz (a16z), Pande left the firm in June of last year to co-found VZVC, a much leaner operation built around a handful of concentrated investments each year rather than dozens.
Pande was once better known in academic circles than investment ones. A Stanford chemistry professor, he gained recognition for creating Folding@home, the distributed-computing project that harnessed millions of home PCs into a supercomputer for disease research. That reputation shifted about a dozen years ago when Marc Andreessen and Ben Horowitz, who had deliberately avoided healthcare and life sciences during the firm’s first five years, reversed course and handed Pande the reins of their new bio and health practice.
His new venture, launched with longtime investor Zach Werner, takes a sharply different approach. VZVC makes only a few concentrated bets annually, employs no associates, and relies heavily on AI for daily operations. As Pande put it, the firm is not doing 30 bets a year.
From a Science of Discovery to Engineering
Pande describes biology as moving from a “science of discovery” toward something that can be engineered. Historically, drug development involved a significant element of chance. What has changed, he says, is that AI and machine learning now let computers grapple with extremely complex problems: identifying which targets drugs should hit for specific diseases, designing those drugs, and even assisting with clinical trials, the most expensive stage of the process.
Cost and time to reach clinical trials have been shrinking, particularly with AI. Even so, running a trial can still cost hundreds of millions of dollars, a major reason drugs remain expensive. Pande notes that the probability of a drug advancing successfully from the first trial through the end of the third is just 20%. When 8 out of 10 candidates fail and each carries a price tag in the hundreds of millions, the amortized cost climbs steeply.
The core reason for those failures, according to Pande, is not human error. Drugs are typically designed and tested on animal models such as mice, which are poor predictors of outcomes in humans. AI models will not be perfect, he argues, but they stand to be considerably more predictive than any animal model, and clearing that bar is where the technology becomes compelling.
The Data Problem in AI-Driven Biotech
One of the harder challenges Pande highlights concerns biological data. Unlike text, which can be scraped from the internet, biological data cannot. As a result, nearly every company ends up constructing its own walled-off dataset. That reality raises pressing questions about who ultimately gains access to the advances AI in medicine has promised, and how broadly those benefits will be distributed.
The next phase Pande points to centers on whether a given drug is the right drug for a specific patient, pushing toward more personalized treatment. VZVC launched in June of last year.
Source
Image: techcrunch.com
