Gene gold turning to dust?
Governments are sinking further billions into genomics and related research but a new study finds no sign of revolution in healthcare.
DNA sequence information can't predict the rich tapestry of life, and researchers are turning to analysing downstream processes using the biotech microarray wonder tool, only to end in disarray Dr. Mae-Wan Ho
Gene microarray studies (Box 1) have been growing exponentially since the mid-1990s. By 2003, thousands of studies were carried out; but that was when things started to unravel.
Microarray for comparing gene transcript
A microarray of short DNA sequences stuck on a glass plate allows two populations of gene transcripts coding for proteins from different cells (e.g., disease versus controls), or the same cells exposed to different conditions, to be compared. One of them is labelled with a green fluorescent dye, the other with a red fluorescent dye.
Spots that appear green are genes expressed preferentially in the green-labelled population; those that appear red are preferentially expressed in the red-labelled population. Those that appear yellow are expressed to the same extent in both populations. The intensity of the colour is proportional to the degree of gene expression.
Margaret Cam, director of DNA Microarray Core at the National Institute of Diabetes and Digestive and Kidney Diseases wanted to use microarrays to study gene expression in pancreas cells. She and her research team used the same RNA samples on DNA microarrays from 3 leading suppliers: Affymetrix, Agilent, and Amersham, and got wildly discordant results. Out of 185 genes common to all three arrays, the expression pattern of only 4 genes agreed with one another. In other words, the noise level could be as high as 98%. The results were in Nucleic Acids Research in 2003.
Marc Salit, a physical chemist at the National Institute of Standards and Technology said Cam's findings caused "one's jaw to drop". Hers was not the first paper to find such inconsistencies. A few ex-enthusiasts think that the promise of gene arrays may have been oversold, especially for diagnostics. Richard Klausner, former director of the National Cancer Institute, now at the Bill and Melinda Gates Foundation in Seattle, Washington, admitted to having been "naïve" to think that new hypothesis about disease would emerge spontaneously from huge files of gene-expression data. The more data he gathered on kidney tumour cells, the less significant they became.
Each company used different short DNA sequence probes spotted onto the array; and they were not telling what exactly these sequences were, so each sequence could be picking up different genes.
Supposedly different probes were responding to pieces of the same gene. Targeting different parts of the same gene can be a problem because genes contain many components that can be spliced into variant mRNAs. The probes have not been designed to be specific to gene-splice variants, and no one has even created a master list of variants for any gene.
Another confounding factor is promiscuous matches. Probes often respond not only to gene products that exactly fit the sequence but also to those that cross-hybridize with near matches. Moreover, many probes don't correspond to the annotated sequences in the public database.
The results from several high-profile papers have already proved difficult to reproduce. Statistician Ulrich Mansmann and his team in the University of Heidelberg pointed out that a series of papers published in high prestige journals like Nature, NEJM, and The Lancet base their impressive results on ad hoc methods, so it is nearly impossible to assess the quality of the studies. They referred to microarray studies as "a methodological wasteland".
"So, despite considerable hype, the published studies are far from the level of evidence that would be accepted for virtually any other medical test." Said the senior editors of PloS Medicine, one of whom, Virginia Barbour is on the advisory board of the Microarray Gene Expression Data Society.
The problem doesn't end there. Many aspects of modulation and regulation of cellular activity cannot be investigated on the level of DNA or RNA transcripts, but require analysis of the proteome (complete profile of proteins). So microarrays of antibodies to proteins have already been contemplated.
Several studies in yeast and higher organisms demonstrated a poor correlation between mRNA and protein, due to a number of additional processes such as posttranscriptional control of protein translation, post-translational modification of proteins, and protein degradation. The current estimate is that there are more than 200 types of protein modification; and that 5-10% of the mammalian genes code for proteins that modify other proteins.
Consequently, the human proteome is expected to range from 100 000 to several million different protein molecules, in striking contrast to the small number of genes. Furthermore, no function is known for more than 75% of the predicted proteins of multicellular organisms, and the dynamic range of protein expression can be as large as 107.
"Knowledge of genomic sequences and transcriptional profiles do not allow a reliable description of actual protein expression, let alone an examination of protein-protein interaction or prediction of the protein's biochemical activities." Said Wlad Kusnezow and Jorg Hoheisel of Functional Genome Analysis in Heidelberg, Germany.
Article first published 23/03/05
Got something to say about this page? Comment