We can now visualize pathogens down to atoms; design vaccines in weeks; manufacture them in microbial factories; engineer them more precise than ever before.
The future holds even greater breakthroughs, but only if we continue to invest in them.
To understand how much vaccine technology has advanced, it’s helpful to take the smallpox vaccines as an example.
When Edward Jenner developed the first vaccine, against the smallpox virus, no one knew what viruses were, let alone connected them to diseases.
Stocks of his vaccine died out repeatedly, and needed to be derived from scratch again and again.
Keeping a vaccine alive in the nineteenth century required arm to arm chains of transmission to preserve the material.
This year @demishassabis predicted AI could cure all disease in a decade.
But Claus Wilke & Derek Lowe say biology is far more complex, or progress will be limited by clinical trials & economics.
In a new 4hr episode of the Hard Drugs podcast, we answer:
Will AI solve medicine and cure all diseases (within a decade)?
We talk about drug discovery, virtual cells, the Human Genome Project, manufacturing, nanobots, innovative clinical trial design, and much more.
AI is already being used in drug discovery, and there’s been a lot of progress predicting the structure of soluble proteins, tweaking proteins and designing new structures, as we’ve covered in previous episodes.
But there’s still a huge gap in understanding protein dynamics and interactions, as there are many areas where measurement tools and data collection are limited, including events that happen in the span of milliseconds or microseconds, which is how fast many things occur in biological systems.
And while computing has scaled exponentially with Moore’s Law, drug development has faced the opposite: Eroom’s Law, where innovation has gotten more complex and more expensive over time.
Even with promising drug candidates, we talk about why human testing – not in animals or virtual cells – will continue to be vital, to test which ones are effective and safe, even though models will help earlier in the pipeline.
Beyond that, large samples and long follow ups are needed to detect rare side effects, understand whether drugs cause long-term complications, and find ways to manage them. It’s hard to see AI getting around the desire for rigorous safety data in real humans.
Another big challenge is the capital and expertise needed to produce and scale personalized medicines and complex biological products, surgeries, transplants, antibodies, and gene-editing tools, which have entirely different cost structures from small molecule drugs.
Their manufacturing and delivery often require highly skilled staff and expensive, intensive, individualized procedures.
Cost challenges are also severe for tropical and rare diseases, where the financial return to diagnose, do research, develop drugs, manufacture and deliver them at scale, is limited.
Without philanthropic funding and economic growth, a lot of diseases are going to remain uncurable, and a lot of people are going to go untreated – whether that’s because of a lack of trust, poor economic and financial incentives, limited public health ambition, and policy.
In one sense, we’re skeptical that AI can solve medicine on its own. But in another, there are many areas where we think AI can help.
So the episode also functions as a roadmap to speed up medical progress and scale up the delivery of lifesaving medicines – with AI and other approaches to reform the pipeline.
What are the economic incentives, innovative trial designs, and data collection efforts that can help drive further medical progress? And how does AI fit in?
You’ll have to listen to find out!
Timestamps:
0:04:34 Contrasting AI optimism and skepticism
0:32:44 The non-linear path between science and technology
1:01:30 The fundamental need for experiments
1:23:15 Animals, organoids, and virtual cells
1:50:47 The challenges of collecting drug efficacy data in humans
2:34:02 The long road to drug safety data
3:06:09 The cost problem of delivering biological drugs and personalized medicine at scale
3:45:35 The global skew in R&D and healthcare funding
4:01:48 Trust, ambition, and the final barriers to medical progress
Listen, watch or read wherever you get your podcasts!
In February, the US government ended some of the most important survey programs globally, used in research & policy for lower- and middle-income countries.
In a new article, I explain why the Demographic and Health Surveys were so critical and why they should be rescued. /1
Hey researchers, bloggers, everyone interested in fertility data!
We recently added lots of new data at Our World in Data on fertility rates, ages at childbirth, twin birth rates, birth seasonality, and more.
Here's a thread of what you can find on the site! 🧵
Our new data comes from a range of sources including the Human Fertility Database, the UN World Population Prospects, the Human Multiple Births Database, the Human Mortality Database, and more.
First, of course, the total fertility rate — a metric that summarizes birth rates across age groups of women in one particular year.
I explore the baby boom in 7 charts, including some trends you (probably) didn't know:
1. Birth rates began to rise in the 1930s, before World War II
The baby boom is typically defined as the years 1946—1964.
For example, Brittanica’s entry states that the baby boom is “the increase in the birth rate between 1946 and 1964”. Similarly, the US Census Bureau defines baby boomers as “those born between 1946 and 1964”.
But as the chart below shows, the rise began earlier.
Birth rates in the US had been falling in the early 20th century. This slowed down and in the late '30s, they turned around & began to rise, which continued during parts of WWII. At the end of WWII, they surged, but this was part of a multi-decadal increase.