Van Eyk, Dunn - Proteomic and Genomic Analysis of Cardiovascular Disease - 2003 (522919), страница 37
Текст из файла (страница 37)
This indicatedthat the gf300 arrays that had been manufactured over a two year period incorporated the same error in printing, and that this error had not been detected by thestandard quality control protocols followed at ResGen.Although still unwilling to sequence all 5147 cDNAs on the array, ResGen waseventually persuaded to sequence a larger subset of cDNAs.
From this larger sample they would determine if there was indeed a larger problem that warranted sequencing the entire clone set used to print the gf300 arrays. The limited set ofcDNAs were sequence-verified from the plates used to print the arrays, and compared to the sequence of the original stock plates and to the expected GenBank sequences. We submitted a list of 181 cDNA clones from the set of 5147 that werespotted onto the gf300 array. The clones we selected were comprised of cDNAsthat appeared interesting to us based on our original data analysis and someothers cDNAs picked from across the array at random. The set of cDNA locationswere submitted to ResGen in July of 2001.The first 90 out of 181 (50%) cDNAs returned to us appeared correct, i.e., theidentity of the cDNA spotted onto the array matched the predicted cDNA sequence as reported in GenBank for that clone.
Furthermore, these clones werespotted at the correct grid location on the gf300 arrays. However, over the courseof the next several weeks, a widespread and disturbing pattern of errors was present in all remaining clones. Of the 181 clones we questioned 91 (50%) had sometype of error. 53 (29%) were found to be mis-identified on the array, meaning thatthe cDNA actually spotted onto the array did not match the expected cDNA in thedatabase for this array.
At 19 (10%) of the locations we queried, two differentcDNAs had been spotted, only one of which was correct. At another 7 (4%) spotlocations, there were two or three different clones, none of which were expectedat that location. Finally, at 12 (6%) spot locations, the cDNAs were unable to besequenced at ResGen due to non-viable bacterial stocks.Some of the cDNAs spotted onto the gf300 array were listed in the set of clonesused to produce the gf301 array, a different microarray product containing a different set of rat cDNAs.
Although we did not check any clones on the gf301 array,the results of our inquiry also calls into question the identity of clones on thegf301 as well. Several of the cDNAs used to generate our arrays were apparentlymis-identified by the cDNA source clone bank prior to being purchased by ResGen for use on their arrays. Unfortunately, the sequences of the entire clone setwas not confirmed independently by ResGen prior to printing the gf300 arrays, sothat errors in the original clone database carried over to the set of clones used forarray printing.Extrapolating to the entire complement of cDNAs spotted onto the gf300 filter, itis possible that 2,265 of the 5,147 cDNAs present on the array may be misidenti-1191207 A Cautionary Talefied and 308 additional clones may be non-verifiable.
To be sure, having an arrayof even 2,573 cDNAs can be very useful for genomics research. The rub is thatone needs to know which loci are reliably identified and which are not. This information is currently not available.7.5Phase 4: Future DirectionsResGen has now agreed on the scope of the problem and has begun a large-scaleeffort to correct the identity and spot locations of the cDNAs present on the gf300array.
Recently, they have discontinued the sale of individual rat cDNA clones andadded some information as to the troubles of “misalignment” of their gf300 arrayson the company web site. However, a general recall of the gf300 rat arrays notbeen undertaken. In fact, ResGen continued to market, produce, and sell thegf300 filter as a “sequence verified” cDNA microarray as of March of 2002, a fullnine months after the problem was identified by us and six months after theproblems were verified by ResGen.
Currently, ResGen is attempting to perform acomplete sequence verification of all 5,147 cDNA clones used to manufacture thegf300 arrays.The final results of our work with the gf300 arrays will not be known untilevery spot is sequence-verified from the stock plates that were used to print them.In cases where two or more cDNAs are present on a single spot, it will be impossible to use the array data for relative expression levels of either of the cDNAs. Weare hopeful that the number of array spots that fall into this category is a smallpercentage of the total.
However, from our limited sample, 14% of the spot locations contain two or three different cDNAs, which extrapolates to over 700 spotson the final array.The experience we have had with microarray analysis of gene expression hasled to several changes in how we are carrying out our experiments. Because largescale rat arrays are not available from other vendors, we have opted to switch tothe mouse and repeat our entire set of experiments. This will permit the use ofour institution’s newly developed array center.
Other issues and ideas are discussed below. We feel that the wider scientific community may have an interest inknowing about this particular array spotting problem as these filters have beenused in published studies of gene expression [4–6].7.6Caveat Emptor: Suggestions for New Array UsersThe following suggestions may help a research scientist who is new to the field ofmicroarray analysis of gene expression to avoid similar problems that can lead toa substantial loss of research time.
Our experience is with cDNA arrays that aremade by spotting individual cDNAs, made from PCR fragments onto a nylon7.6 Caveat Emptor: Suggestions for New Array Usersmembrane. The source of our troubles stemmed from a mix-up of the cDNA clonesprior to printing the arrays. Oligonucleotide based arrays will not have this type ofproblem, because the oligos are synthesized directly on the surface of the chip. However, the curation of DNA sequences used to design oligo based arrays is just as important as the curation of individual cDNA clones that are spotted onto an array. Arecent recall of arrays manufactured by Affymetrix, Inc. was due to errors in the publicly archived genetic information that went into producing the chips, eventuallyleading to mistakes in the oligo design used for these chips [20, 21].As with any scientific experiment, when planning experiments using microarrays it is important to control for experimental, biological, and technical errors.With many arrays having over 10,000 individual cDNA spots, it is impossible toindependently verify each result without a gargantuan expenditure of resourcesand time.
The point of having an array of cDNA spots in a small area is to bringthe possibility of examining whole transcriptomes from a cell or tissue at once.For this reason, it is vitally important that the array manufacturer takes the necessary steps to ensure the identity of every sequence and the location of each ontheir arrays. It is equally important that the user understands the specific qualitycontrol and testing procedures that are utilized in the array production process. Inthis respect, purchasing microarrays, even from an in-house distributor based atyour university or within your company, is different than purchasing a particularchemical reagent where it would be possible to verify the quality of the reagentwith a simple test or two, such as checking the pH.
With microarrays, it is notpractical for you to do any extensive quality control yourself so you must be confident that the manufacturer is being exceedingly careful of all aspects of microarray fabrication. In our own experience, this sadly was not the case.Based on our experience, we believe it is of the utmost importance to have atleast the following information prior to investing the substantial time and moneyrequired to start using array technology. While it is often difficult to obtain thespecifics from an array manufacturer, one should demand as much informationas possible. Avoid doing business with a vendor that does not give you concise, accurate, and believable answers to the following.
These questions are best answered by the director of quality control.1. How were the cDNAs prepared?It is standard practice to use multi-well plates to maintain individual cDNAsin the form of plasmids or PCR products that are used to print out spots onnylon filters or glass slides, although the specific steps may vary from manufacturer to manufacturer. Manipulations of liquids are carried out by a roboticinstrument that physically interacts with each sample, repetitively transferringa portion of the original cDNA or glycerol stock from each well to a replicateplate that serves as a template for a PCR reaction. The replica plate is cycledthrough standard PCR conditions and the resulting PCR products are pickedup by a robot and spotted onto the surface of the array.
There are several stepsduring the array printing process that can lead to spotting errors. The individual PCR products can become cross-contaminated with one another during1211227 A Cautionary Tale2.3.4.5.6.7.the PCR process. This could be due to misalignment of the individual template plates, bent pins on the robotic transfer head, or incomplete washing ofthe pins between cycles. Good quality control practices will reveal the existence of either of these problems, and it is standard practice to check the integrity of the pins.
![](https://s.studizba.com/z.php?f=/templates/si/i/noavatar.png&w=100&h=100&t=1"){kind=avatar}
