For 30 years, it has been known that testing hypotheses based on a normal distribution and a 95% confidence interval is insufficient. A 99% confidence level is impossible to achieve using experimental methods.
Bayesian probabilities have been used to some extent but have not really caught on in the scientific community.
It has been estimated that only 30% of published experiments in cancer research are reproducible. Nature, other journals, and institutions have joined forces to secure funding and reproduce the experiments. Without much success.
I regularly consult the Cochrane Library. Typical meta-analyses range from “not enough results” to “out of 482 studies, we selected 5 as valid, 4 of which were by the same author” (I can’t remember if it was on pregnenolone, melatonin, or something like that).
If you've only ever studied math, this creates cognitive dissonance for you (if it's any consolation, you're not the first mathematician I've met in this situation).
I recommend that you explore epistemology, starting with Thomas Kuhn
https://en.wikipedia.org/wiki/The_Structure_of_Scientific_Revolutions
and to continue with Popper, Bachelard, Habermas, and Feyerabend.
At some point, you will realize that it is your view of science that limits you, not the way science is actually done.
The word bio-logy contains the Greek word logos; this science has no other claim than to
discuss what we understand about living things. Even the definition of “life” remains unsatisfactory.