#485 - The New Science of Cancer

The gist

Siddhartha Mukherjee, author of the updated Emperor of All Maladies, discusses how our understanding of cancer has fundamentally shifted over the past 15 years. Rather than one disease, cancer is best understood as thousands of individual mutations requiring personalized approaches. The conversation spans three critical areas: prevention through newly discovered "inflammagens" that alter cancer's microenvironment, the pitfalls of premature detection via liquid biopsies (hampered by Bayes' theorem), and treatment breakthroughs that have transformed certain cancers from death sentences into manageable chronic conditions.

Key takeaways

• Cancer is genetically individual: Each patient's tumor has a unique mutation profile. Two women with "breast cancer" may have completely different genetic drivers, requiring different treatments—there won't be one cure but hundreds.

• Prevention's missing biomarker problem: Unlike heart disease (which has cholesterol and blood pressure as surrogate markers), cancer lacks a reliable biomarker for future incidence, making prevention trials extraordinarily long and expensive. No major preventable chemical carcinogen has been discovered since the 1960s.

• "Inflammagens" represent a new prevention frontier: A new class of carcinogens (like particulate air pollution and asbestos) doesn't cause mutations directly but creates chronic inflammation in the tumor's microenvironment, awakening dormant cancer cells. Identifying and measuring this specific inflammation could unlock chemoprevention.

• Bayes' theorem explains liquid biopsy false positives: A positive cell-free DNA test feels reassuring at "99.5% specificity," but when cancer's base rate in healthy populations is low, most positive results are false positives. The test's utility depends entirely on prior probability—it's most valuable in high-risk populations (prior cancer, genetic mutations, strong family history).

• CAR-T therapy works for blood cancers but struggles with solid tumors: CAR-T cells successfully treat leukemias and lymphomas but cannot penetrate the protective microenvironment ("soil") surrounding solid tumors, limiting current applications.

• Immunotherapy and targeted drugs have delivered real cures: Non-small cell lung cancer now shows 20% five-year survival rates (previously considered incurable); myeloma survival improves every five years; childhood ALL cure rates exceed 90%; and new RAS inhibitors extended pancreatic cancer survival from 6 to 13 months—the first meaningful progress in decades.

• AI is transforming drug discovery by teaching machines medicinal chemistry rules: Mukherjee's company, Manus AI, teaches AI the rules of drug chemistry rather than relying on insufficient training data, enabling more efficient molecular design and target discovery.

• HPV vaccines are a complete cancer prevention success: A Swedish randomized trial showed that appropriate HPV vaccination reduces cervical cancer risk to virtually zero—this is a completely preventable cancer, yet thousands of women still die globally due to vaccine hesitancy.

Notable quotes

• "Every individual form of cancer, every individual specimen of cancer is its own disease in the genetic sense." — Siddhartha Mukherjee

• "The big unlock for treatment is always going to be: can we find something in the cancer cell that's different from the normal cell? Cancer cells are very close cousins to normal cells, because they're derived from normal cells." — Mukherjee

• "Thomas Bayes knew the answer two hundred odd years ago...It's amazing to me that in 2026 we're having this anxiety-ridden conversation about whether to test, when the answer is: What is your prior probability?" — Mukherjee

Worth a full listen?

Listen to the full episode if you're navigating personal cancer risk assessment, considering screening tests, or want to understand the scientific and economic foundations of modern oncology—Mukherjee's explanations of Bayesian reasoning and prevention science are unusually clear and personally relevant; the TLDR captures the main arguments but misses valuable context on how to apply this framework to individual decision-making.