[nonbel]

Its just an excuse, you are talking about a field with 10% replication rate, and when they tried to start doing more replications it was too expensive to even figure out what generated the data so they had to drop 25% of the studies before even beginning. One guy even calling the literature "augean stables":

https://www.nature.com/news/cancer-reproducibility-project-s...

Every disease can be arbitrarily split or combined into multiple categories. It doesnt mean anything. Here is a cure for cancer:

1) Detect aneuploid cells that fail to show the initial steps of the response to extracellular apoptosis signals, and have been dividing more often than usual for that tissue/environment (eg you have to take into account wound healing, etc).

2) Kill those cells.

This is technological problem, not a "cancer is many diseases" problem. Every problem can manifest in many ways and have multiple solutions.

EDIT:

Ref for "10% replication rate":

>"scientific findings were confirmed in only 6 (11%) cases." https://www.nature.com/articles/483531a

More info in this (short) thread from a few years ago: https://news.ycombinator.com/item?id=10687879

[jmcgough]

I know a lot of smart people who've devoted their lives to solving this problem. It's very easy for you to dismiss it as easy or say that researchers are incompetent when you haven't spent years working on cancer research in a wet lab.

[nonbel]

I have worked in a "wet lab" for years, but not cancer research. The first 3 were largely (of course there is always learning going on) a waste because I didn't understand what statistical significance really meant. I can't really blame myself since neither did anyone else in my committee, department, or seemingly the entire area of research. But anyway this meant the whole project was misdesigned from the start (the design was standard).

The next two were spent attempting to save the project by working 18 hour days almost constantly. In the end, despite what they claimed about doing it the usual way because "its so complicated, no one can come up with mathematical/computational models or actual predictions to test, no one can describe this or that in so much detail, etc", the actual truth was that it wasnt "too hard". It was that no one really cared about that. Instead they just wanted to know whether there were statistically significant differences anyway...

From what I've read of the cancer literature it looks exactly the same, probably worse. And guess what, there's been no real progress made for decades on the topic I was studying either.

[nonbel]

I see now that they had to scale down again to only 36% of the original results:

>"A large effort to reproduce high-impact cancer research has scaled down the number of studies it plans to replicate from 50 to 18, Science reported yesterday (July 31).

[...]

Hurdles to replicating experiments included a lack of detailed protocols and easily obtainable reagents. “Communication and sharing are low-hanging fruit that we can work on to improve,” Elizabeth Iorns, the president of Science Exchange, tells Science." https://www.the-scientist.com/news-opinion/effort-to-reprodu...

Yup, the entire system is set up for an environment of no replication. How do you expect to cure cancer without making reproducibility a priority? I'd expect people who don't do that will end up thinking everything is "so complicated" due to all the conflicting results they generate, eg "one disease" becomes "many diseases".