Nature杂志上的实验设计经验综述
Nature Reviews Molecular Cell Biology 5, 577-581 (2004);EXCESSIVE TRUST IN AUTHORITIES AND ITS INFLUENCE ON EXPERIMENTAL DESIGN
Tung-Tien Sun
Tung-Tien Sun is at the Epithelial Biology Unit, Departments of Dermatology, Pharmacology and Urology, New York University Cancer Institute, New York University Medical School, 550 First Avenue, New York 10016, USA. [email]sunt01@med.nyu.edu[/email]
Most graduate curricula in biological sciences offer courses that cover various scientific disciplines, but they give relatively little formal instruction in experimental design. Students learn the latter primarily through hands-on experience in the laboratory, and some find this learning process bewildering and frustrating. So, what is the root of the problem, and how can young researchers get experiments to work more predictably and reproducibly?
Central to biological and other experimental sciences is the practice of formulating a hypothesis and subjecting it to vigorous experimental testing. Most graduate schools teach students how to survey the scientific literature and emphasize the scientific paradigms that define important and solvable problems1, and therefore provide a basis on which to formulate hypotheses. However, few graduate schools offer formal instruction on how to become a good experimentalist. Our focus on the intellectual aspects of research (asking the right questions) is completely appropriate, but it should always be accompanied by a solid training in laboratory investigation (getting high-quality data). Traditionally, we place our students in laboratories and hope that they will learn all there is to know about how to design and carry out good laboratory experiments from the example of senior investigators and colleagues. Although many students function well under such an informal, and rather variable, apprentice system and become good experimentalists, some new students find this learning process bewildering and frustrating. This is exemplified by the following comments that I have heard. One student said, "I asked several senior graduate students and postdocs how to do an experiment. But everybody gave me a different answer. Whose advice should I trust?". Another said, "I followed a postdoc's detailed written protocol to the letter and I even used all his solutions that worked for him. But my experiment failed and I lost three months. I am really upset!".
So, what is the root of the problem, and is there anything we can do to help these students? I propose that excessive trust in authorities is a key underlying cause of many experimental failures. In this article, I discuss how an awareness of this problem and a few practical considerations, including the concept of risk assessment, can help to get most laboratory techniques to work more predictably and reproducibly. Many of the practical tips that I discuss are so instinctive and logical that most experienced scientists practice them naturally. However, I hope that young investigators will find these messages helpful and reassuring. Finally, although most of my comments are directed towards the design of biochemical experiments, they are generally applicable, with minor variations, to other experimental sciences (for further useful information, please refer to Refs 2–7).
Trust in authorities and risk assessment
The American Heritage Dictionary defines authority as "an accepted source of expert information or advice", and there are many elements in an experiment that might be regarded, sometimes uncritically, by a new experimenter as an 'authority'. These include professors' or other experts' opinions, books and published experimental protocols, labels on reagent bottles, and instrument readouts. When any of these are perceived as 'authorities', it means that the experimenter has decided that nothing can go wrong with such information or reagents. Accepting many (and sometimes all) elements as authorities simplifies a person's thinking, is convenient and gives a sense of security ("Well, the experts said so". But each of these authorities constitutes a blind spot in our experimental design: when an experiment fails, we tend to overlook these elements as potential causes of the problem, and this compromises our ability to troubleshoot.
Although some purists might say that scientists should never trust any authority at any time, when carrying out experiments we, for practical reasons, invariably choose to trust some of the elements. The increased reliance on commercial experimental kits that frequently contain undisclosed, proprietary formulas, together with a less rigorous background in chemistry and other physical sciences, underpins a worrisome recent trend in which more of our students seem willing to trust authorities, sometimes excessively ("As long as the kit works, why should I care how?". In this article, I emphasize the importance of deciding how much we can trust certain elements as authorities under a given circumstance. A practical question that we can ask ourselves is, how much can I afford to make (potentially preventable) mistakes in this experiment? For an experiment that takes only a day or two and uses inexpensive reagents, we might decide that we can afford to make such mistakes. But, for experiments that take months to do and/or require expensive or irreplaceable materials, we cannot afford to make potentially preventable (also known as stupid) mistakes. For example, the generation of transgenic or knockout animals takes months before we know the outcome, so it would be foolish to use casually characterized complementary DNAs or genes for this purpose. Another 'experiment' in which no preventable mistakes can be tolerated is the space-shuttle project. Although it has failed several times despite all humanly possible efforts, Gene Kranz — a former NASA flight director — summed up the attitude of NASA workers by saying, "failure is not an option" (the title of his 2000 New York Times bestseller book). It is therefore important to assess risk before carrying out every experiment: the outcome of this analysis (risk assessment) determines how much we can afford to blindly trust various elements that might be regarded as authorities (risk management; Fig. 1).
Figure 1 | The importance of planning in experimental design.
This figure outlines a decision-making process that should take place before every experiment, and it emphasizes the importance of detailed planning and thorough understanding in experimental design. Rushing into experiments without thoughtful planning invites failure. Perhaps this is why someone once said, "Seventy percent of whether your experiment will work is determined before you touch the first test tube.".
Failure in risk assessment
In addition to the opinions of professors, so-called 'experts' and textbook authors, who are clearly not infallible, commercial reagents and materials can also be a problem. For example, a young mouse geneticist was hired by a well-known university some years ago, and was given a laboratory and supervised several graduate students. The geneticist ordered batches of mice from a leading animal supplier to carry out breeding experiments. After several years of hard work, it was discovered that some of the supposedly inbred mice were impure or of the wrong strain. All of the geneticist's painstaking and time-consuming work amounted to nothing, and led to no publications at the time of tenure review. The geneticist sued the animal supplier, but was told that the animals were shipped with a warranty that essentially stated, "This warranty limits our liability for replacement of the product. No other warranty of any kind, including fitness for a particular purpose, is provided.".
页:
[1]
