The human body is made up of molecules that fit generally into the categories of DNA, RNA, proteins, sugars, salts, fats and water. These last four groups are generally governed by the proteins in the human body and become imbalanced as a result of protein dysfunction. Technological strategies, developed as a result of the Human Genome Project, can rapidly and almost comprehensively scan through the DNA, RNA and protein molecules of the human body in order to identify differences between individuals with a disorder versus those without. These strategies are collectively known as the “-omics.”
Transcriptomics refers to the comprehensive scanning of the nearly fifty thousand currently known genes that are transcribed into RNA molecules from the three-billion-letter human genome. Each cell utilizes (expresses) different genes at different times in its development and under different physiological conditions. In general, tissues express similar sets of genes that can be used to identify those tissues in the absence of any other information. For example, the brain expresses about thirty percent of all of the known genes; those specific transcripts are different from the transcribed genome in the heart. We can therefore define molecular signatures based on expression profiles, and these profiles can then be used to automatically separate normal cells or tissues into their correct category.
This method of expression profiling can also be done with disease states and implemented diagnostic tests. For example, neurons that have one of the beginning hallmark features of Alzheimer’s disease, neurofibrillary tangles, show altered gene expression patterns when compared to normal neurons. This information can be used as a molecular diagnostic in the absence of histopathology for tangles because it is, in fact, a surrogate marker of that cellular condition.
The same can be done for clinical states that may be less obvious or for which there does not exist a gold-standard diagnostic. For example, in autism we are defining the expression correlates of the disorder by sifting through the entire transcriptome of hundreds of individuals with the disorder versus hundreds without the disorder. Each individual has undergone ten hours of clinical assessment and has either been diagnosed with autism or determined to be unaffected. Once the expression signatures that are specific to autism are established and validated (through blindly diagnosing several more hundred individuals), we can then use the expression profile as the diagnostic test in order to circumvent the ten hours of clinical assessment (contingent on the accuracy of the test, of course). We could also potentially begin testing in a pre-symptomatic scenario to encourage early interventions and result in better outcomes.
Another example of the power of transcriptomics is its ability to sub-classify disorders that on the surface appear to be similar. For example, drawing on the autism example above, we can identify molecular sub-classes of autism that correlate to differences in the clinical picture. For example, many children (but not all) have gastrointestinal symptoms. Some children respond to dietary modifications, but there is a large spectrum of behavioral symptom severity. By expression profiling the entire fifty thousand transcripts in a cohort of children with autism and then performing computational analyses (which allows the profiles of molecularly similar children to congregate next to one another), we can create similarity dendograms that define molecular subclasses. All available clinical data can then be overlaid and permuted according to molecular subclass to see if correlations exist. If a correlation does exist, then that profile can be used as a diagnostic test for that clinical parameter. This strategy has been shown elegantly for a number of disordersâ€”primarily in oncology, where one can define outcome and survival rates and drug response rates based on an expression profile.
In summary, extensively validated expression correlates can be used to diagnose a disease as well as diagnose clinically relevant subclasses. There are a variety of technologies on the market today that can perform the discovery portions of expression profiling (defining the expression signatures). These include vendors such as Agilent, Affymetrix, Operon and others. For a more comprehensive resource on what technologies are available, how to use these technologies and examples of successful discoveries and implementations from the transcriptomics field, please visit the NIH Neuroscience Microarray Consortium web site.