Conclusion
In conclusion, DADA is a digital method for analyzing expression levels based on the counting of restriction fingerprints of cDNA clones. It is built around a specifically designed cloning vector that suppresses misleading fingerprints derived from partial cDNAs or empty vectors. The method yields quantitative data and absolute figures that do not depend on amplification by polymerase chain reaction. Compared to the sequencing of cDNA libraries (conventional EST approach [14]), it is substantially faster and more economical. In comparison to the SAGE method [20] it results in physical cDNA clones of substantial length which speed up further analysis.
Within the limited number of examples, there is a good correlation between the clear results from DADA and other gene expression analysis methods such as quantitative real time RT-PCR, RNase Protection Assays, and various hybridization procedures (data not shown). A number of additional genes with differential regulation discovered by DADA and confirmed by independent methods are now under investigations for the use as therapeutical targets (unpublished observations). Some of the genes identified by DADA including MCP-2 and Cystatin C were neither discovered by means of subtractive hybridization [3], nor by differential display [9], although these methods were successfully used in comparable settings in our laboratory and lead to a high number of differentially expressed genes. Vice versa, DADA could not detect certain differentially expressed genes, that were discovered by standard methods. As an example, the injury-induced differential expression of S100A9 could be detected by subtractive hybridization [33]. The DADA fingerprint of S100A9 was detected in both analyzed cDNA libraries. However, as it occurs only once in each library, no statistical significant differential expression could be identified by the DADA method. Only a subset of genes was identified by all methods. In general, the different approaches can be seen as complementary rather than competitive. The specific findings will certainly vary with changes in the completeness and individual execution of the respective screens. Additionally, DADA offers advantages with regard to absolute and quantitative data at the level of screening.
As of yet we can only perform digital analyses with patterns experimentally analyzed from single clone analysis. These fragment lengths reproduce accurately enough to search and count them in mixed analyses applying rigid queries. The usage of restriction fingerprints computationally derived from sequence database entries is hampered by the lack of predictability of gel positions from sequences and the integer versus non-integer issue. The experimental fingerprints, due to the narrower search space, allowed for the identification of fingerprints in mixtures of 96 colonies as demonstrated here. The broader search space of sequence-based predictions allow only for the identification of fingerprints in mixtures of about 10 colonies (data not shown). The one order of magnitude lower throughput in the case of predicted fingerprints would render the method non-superior to others.
The generation of experimental fingerprints is the bottleneck of the procedure as described here. Once the fingerprints are generated for a certain species they can be rapidly applied in all possible settings. Therefore, we are developing improvements of this step. In addition, the use of a now available fifth fluorescent dye as a size marker in sequencing lanes should allow for a fast correlation between restriction fragment sizes (derived from the sequence) and gel run behaviors of DNA fragments (relative to the size marker). The latter improvement could not only lead to a faster generation of experimental fingerprints, but could also supply a large dataset to improve the prediction of gel run behaviors of fragments based on the sequence composition.
Another improvement could result from alternative DNA separation and detection methods such as capillary electrophoresis or mass spectrometry. This could further improve the reproducibility of the fragment analysis and lead to a reduction of the search space for every fragment. As discussed above, this parameter is directly linked to the throughput of the method. In addition, DADA would profit from the higher throughput of capillary electrophoresis compared to the gel electrophoresis used in this study. On the other hand, values created by mass spectrometry are precisely predictable from the sequence of DNA fragments [29]. This would allow for the generation of fingerprints from sequence data in silico and speed up this limiting step of the procedure.