I got requests for commentating and revising research proposals all the time, mostly from trainees and junior PIs. However, I found a serious problem: many of them did not know how to write a hypothesis. They did not understand scientific method and, especially, the meaning of falsibility.

The beginner's level of mistake is to regard hypothesis as a re-statement of a study aim. For example, "we hypothesize that analysis of gene expression can reveal therapeutic targets." A hypothesis needs to be a theory-deduced prediction that can be tested experimentally. This is a key part of scientific method, which is frequently explained as the following procedure: (1) defining a "why" or "what" question; (2)  constructing a theory to provide an answer to the question; (3) deducing a prediction from the theory; and (4) testing the prediction by an experimental study.

Therefore, a hypothesis should be generated from a theory and predict an unknown but testable phenomenon. In other words, the hypothesis needs to be sufficient to instruct what kind of test needs to be performed. For example, a theory can be as simple as "DNA is the genetic material of cells.", to answer the question "what is genetic material composed of?". We can deduce a hypothesis from this theory: transferring the DNA from bacterial strain A into strain B will make the later to get the phenotype of the former. This was tested by British bacteriologist Frederick Griffith in 1928, validating the cellular function of DNA.    

The renowned philosopher Karl Popper defined the "testable" feature as "falsibility" of a  scientific theory. That is, a theory is scientific only when it provide the possibility to be proved wrong. Therefore, if a theory is claimed always right and provide no approach to test if it's wrong, it is not a scientific theory. For example, astrological theory drives people to find facts to match its prediction, so it is always right and therefore not scientific.    

Popper invented the concept of falsibility for the purpose to distinguish science from pseudoscience, and better theory from worse theory (e.g. Copernican model vs. Ptolemaic model). However, many people mistake that falsibility is the ONLY feature of scientific method. In such idea, as long as the hypothesis is testable in the format, it is a legit one. We can often find in a research proposal, the hypothesis is generated without the base theory and the process of deduction. For example, I've seen many like "compound X can kill cancer cells, so X can be used as a anti-cancer therapy". I am pretty sure adding salt into culture dish can kill many different kinds of cells, including cancer cells. Unfortunately, salt is never used to treat cancer patients. Again, the problem here is the lack of theory and deduction of prediction. In this example, the question could be "what kind of compound can be used to treat cancer?" The theory could be "compounds specifically killing cancer cells instead of normal cells"; so the theory needs to be further developed to instruct how to find such compounds, allowing to deduce the hypothesis. Though looked way over-simplified, this example is actually the "magic bullet" concept developed by Paul Ehrlich in 1907, and eventually evolved to the concept of targeted therapy.

In the history of scientific research, there are abundant examples of well-defined questions, development of theories, generation of hypotheses, and instruction of experimental design to test the hypothesis. We can learn scientific method from them, write a well-formed hypothesis in our research proposal, and make a good study design to test it. Unfortunately, I hardly find the history of (biomedical) research in any curriculum of undergraduate or graduate programs. I sincerely hope that the agencies of scientific education consider it.