Search This Blog

Wednesday, August 09, 2006

Another Case

In my last post I presented the reader with an actual case of scientific misconduct. This kind of misconduct occurs everyday in industry and academia throughout the land. It is the kind of dishonesty where you get to select what data gets used. It happened to the doctor I spoke about who worked with Actonel data. It happened around me at a previous company involving siRNA and TNF alpha. I wanted to open up a discussion with other people who may be feeling guilty about the work they do. How many others put up with this sort of data mishandling? How many actually orchestrate it?

I believe I could come up with a case of similar misconduct from every position I've had except for one. That is to say, I think it's the most common form of Cargo Cult Science. People have lives to lead. They earn the money for those lives by working in science. If you run an experiment 10 times and only once do you get the desired results, you may disregard the other 9 experiments. You may choose to not report them. No one will find out most likely so you keep you mouth shut. The scientist who manufactures data most likely ran his or her experiments more than 10 times but never got the desired results.

Imagine however, if you truly could clone a stem cell. Imagine if you got the desired results because you were right about something. You know what? That still doesn't make you a scientist. You got lucky. Getting the desired results comes from standing on a large volume of information that lead you to design and conduct a proper experiment. Maybe you failed 900 times but each time you learned something. Each time you got closer to the truth. This is real science. Failing is good if you are prepared for it. If not you will be fabricating or selecting the data you want. Either way, you are a fraud.

This next case describes a way around selective data. It involved a protein purification. The goal was to purify a protein to make an antibody. Since there was no antibody available I used a few basic techniques. In the end I had 3 bands on my gel, which to some indicates that I have 3 proteins. What I did showed that there were at least 4. I would guess many more.

The 3 bands were all very close together indicating that there were three major sizes of protein purified. To further separate the bands I ran them on a higher percentage gel. They were still too close together. The cubicle scientist needed one of the bands to be the protein we were after. He merely wanted to be able to point to which band on the gel was the one. He wanted to show the gel and then a western blot of the same gel. The western blot would indicate which band was our protein but the bands were still too close together.

Out of the blue, a passing technician solved the problem. Our protein was glycosylated. Simply run a PNGF digestion, which would deglycosylate the protein, and run the gel and western again. This time the 3 bands showed up at the same location on the regular gel. On the western however the band shifted down well below the three major bands. This meant that none of the bands represented the protein we were after. Furthermore, there was no band on the gel where the western blot band was located. That means there was not enough of the protein to be visible on the gel.

Now the western blot varies in sensitivity. This particular western was very sensitive. As a young technician who had done at least 100 of these westerns by this time, I felt that the 3 bands were too big to give the kind of western results we were getting. The western would have showed a huge black blob up and down the western. Instead it gave a neat little band. What I learned from this was that many proteins migrate with others there size in a gel. A band is probably an aggregate of proteins close to each other in size. The amount of proteins it takes to make a visable band on a gel can be hundred and thousands of times more than you would need to get a nice sharp band on a western. Indeed I learned a lot from that experiment and how to look at these gels that so many use in there research. Each protein antibody combination differs when it comes to western blots. There is a cut off level where slight changes in concentration either way will give big differences in band intensity. At some point they'll remain the same over a range of concentrations. At some point they will give high background signals. Some combinations will detect the protein all throughout the lane but show a blob where most of the protein is.

I also learned that this experiment was not described by the cubicle scientst. He insisted that I give an answer as to the percentage of purity I had acheived in spite of the fact that we both had access to the whole set of data. I came up with less than one percent as my answer. He reported 20 percent. To back this up he showed the gel with 3 bands on it. He showed the western blots and said that it could not be determined which band was the protein of interest but the western proved that it was one of the three. They were just too close together to discern which one it was.

So he lied. He did not manufacture data but he did select what he needed. Furthermore, he left out the data that proved his conclusion to be wrong. I hung my head but I was powerless to change his report. I was out of the picture by now and just a lowly technician. Oddly enough, when the antibodies started showing up we found a few that were specific to the protein. It didn't matter that 99 percent of what I sent out was not the protein of interest. What little there was gave rise to specific antibodies to our protein. No harm not foul right? No one will ever know what happened.

No comments: