The Nobel Prize in Physics 2009 was divided, one half awarded to Charles Kuen Kao "for groundbreaking achievements concerning the transmission of light in fibers for optical communication", the other half jointly to Willard S. Boyle and George E. Smith "for the invention of an imaging semiconductor circuit – the CCD sensor".
The image from a CCD is black and white, but by placing a red, green or blue colored filter over the top of each pixel, color information can be read directly from each pixel — but only for one primary color per pixel. Subsequently, software can also extrapolate the color of adjacent pixels based on their brightness, so that each pixel winds up with its own red, green and blue values. A RAW file is the “raw” color data from the chip before any of the post-processing extrapolation has been done. Cameras usually do all of this processing for you and spit out the result as a JPEG. With the RAW file you actually have all the original sensor data, which is much more information-rich.
So far so good. Here we pack up the Allied forces airport and the Cargo Cult takes over. Next we introduce the western blot. Somewhere between running the western and generating the image of that western we enter the cult.
Post-processing extrapolation was my job for five years. I was a photographer. The original data that I used was a roll of film. I learned how to expose film using f-stops and shutter speed. I learned how to process and print black and white, color and slide film. Interesting things to know but digital technology now takes care of all the things I was doing manually. Post-processing image quality issues like color, contrast, and density are now automatic.
My first few years running western blots were done on film. Then one day we received our first AlphaImager. It is a box with a light table that fold down so you can take pictures of SDS PAGE gels. Fold the light table up and there is a UV light box for DNA gels. Shut off all light and you can take pictures of chemiluminescent western blots. A a RAW file is generated, with all the original sensor data, which is much more information-rich than a processed image.
The western blot has seen some method advancement. There are E-gels and I-blots which speed up the time to transfer the proteins onto the membranes. There are more sensitive reagents for developing the westerns so lower amounts of protein can be detected. Each change challenges the Cargo Cult tribesmen. If a transfer is sped up, they claim, you will get less protein transferring. "How do you know?", I would ask.
Using film, I was told, was superior to digital images when it came time to expose for weak signals. "Can I see the evidence?" I was respond. I wasn't getting a satisfactory answer. I developed a test. I loaded a gel with high to low amounts of protein. The low end concentrations dropped below the level of detection even at maximum exposure time (chemiluminescent probes lose their energy after about one hour). I prepare two identical westerns. One western was visualized by exposing it to film for one hour and the other was visualized using an AlphaImager CCD camera.
Test number one: Hold film in hand and use eyes to compare with digital image on computer screen. Oh the varying opinions! Based on ones bias, a subjective argument would break out. "The film is more sensitive!"
Test number two: Use ImageQuant software from GE Healthcare to quantitate sensitivity. The bias was still there but the data was irrefutable.
Science and technology wins the day.
Next I ran the same western blot and probed it with a fluorescent dye conjugated antibody. I used the CCD camera to capture the image. I also captured the image of the very same western on a lazer scanner. The fluorescent probe provided greater sensitivity measured by the signal from the lowest detectable amount of protein. The scanner produced cleaner data as measured by the correlation coefficient of the linear curve generated by the protein signals. I laid the western in a drawer and let it dry out. Every month for six months I scanned the western. The results did not change.
So I had determined empirically the best way to run one particular western blot. I'm not done yet however. If I am to get reproducible results from my western I must also account for the protein sample I am loading. Purity and stability, for example, must be known. In a proper research organization all of this is known by the QC group. The people who develop the methods take these factors into account. The people who then run the western blots must adhere to the standard operating procedure. They are tracked electronically so that deviations can be accounted for. For example, if a standard curve appears to be lacking in linearity or range, the protein standards can be identified to see if they were used beyond their stability window.
This all leads to the quality control of a cargo cult. I consider a peer reviewed journal to be part of the cult. The tracking of source material, equipment, and technician is not something that a journal editor asks for. It is even rare for immediate supervisors to read over a laboratory notebook, especially when the results turn out in their favor. The cargo cults do not have quality control mechanisms. They rely on their intuition and intelligence. As a result, any questioning of the finalized data is frowned upon. Lower ranking members of the organization are then held accountable for any transgressions, since the scientist above them has such a high standard for their honesty.
If the JPEG or TIFF files were made available we wouldn't have to use our eyes to try and discover fraud. Variances in density, for example, could indicate cutting and pasting. That clear line that gets erased with density adjustment remains on the pre-processed image file. The pre-processed data is what we need, not the post-processed product offered up for publication. The post processing is where the scoundrel plies his/her understanding of imaging and ways of fooling the older generation of leaders. Fortunately, post-processing is tracked by the software. If post-processing takes place, that image must be saved as a new file. The family tree of an image is stored automatically. It all leads back to the pre-processed image.
This is a critical point in the Cargo Cults use of imagery in their ceremonies. The real wealth of the imaging systems comes from the pre-processed data. Office dwelling scientists don't publish relevant information on image enhancements. They offer up aesthetically pleasing images that support their narrative. Orthogonal methods should be required to back up the western analysis, such as the ForteBio.
No comments:
Post a Comment