🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

RCTs cannot be hacked in the same way that biomedical research can be. The fact that the clinical trial success rate has not changed in 50 years tells the real story of how much we actually understand biology.

When you’re a CEO/salesperson and you’re raising hype for your company, hyperbolic predictions will only benefit and never harm you. Might as well just say the craziest thing you can imagine.

The greater the number of data points per measurement, the easier it is to make spurious associations. This is why high-dimensional measurements (of e.g. the microbiome, epigenome, genome) can be linked to virtually anything, to the point of being meaningless.

The fact that high-dimensional measurements can feasibly be linked to anything and everything makes them a prime story-telling tool, even if it lowers their scientific validity. The entire field of genomics has learned to take advantage of this fact.

The figure in this article illustrates exactly how most biology papers are secretly p-hacked. A large number of hypotheses is explored, and only the ones that form a coherent story are reported. This is actually the main reason behind the replication crisis in biology IMO.

It’s quite revealing that the academics who wrote this article are oblivious to the bad implications of this figure. It’s presented as a good thing.

There's a sweet spot to scientific storytelling - your story needs to be EXACTLY 'slightly more complicated than what your audience can immediately grok'. Too understandable and you don't signal expertise and lose mystique. Too complicated and you totally lose your audience.

If your hypothesis is compelling enough, your data doesn’t really even matter.

Drug development was figured out in the 1960s. The field has only been regressing since then.

Biology’s sub-25% replication rate is an EXISTENTIAL issue for the field. Most academics don’t seem to care, which shows that their priorities are not in doing good science.

Two things can simultaneously be true: 1. In absolute terms, we’ve made tremendous progress in understanding biological systems in the past 50 years. 2. Relative to the full complexity of these systems, our current understanding is approximately 0%.

Despite decades of biomedical research, trillions of dollars in R&D, and dozens of Nobel prizes, drugs developed today have the same chance of curing disease in clinical trials as drugs did in the 1960s, when the strategy was basically pure trial and error.

The fact that so many different compounds “cure cancer” in cancer model systems says much more about the model systems than it does about the compounds. Wittgenstein’s ruler.

Having a poor model of the world is often worse than having no model at all. This is because models reduce flexibility, and in complicated/unpredictable domains, the best strategy to make progress is usually to maximize flexibility.

When you tell a scientific story, people are convinced by your choice of experiments, not by your actual results/data.