This is a big picture question for research historians among us. I'm wondering whether any gene expression microarray or RNA-seq study (or series of studies) has identified anything (a biomarker of some sort) that later became *clinically used.*
I work with transcriptomic classfiers like support vector machine and random forests. One of the critiques on a recent grant proposal was this:
Can you cite an example of an instance where transcriptomic studies or classification tools identified anything that became a clinically useful "biomarker"? (For example, used to diagnose, subtype or prognosticate disease, or select the best treatment)
Now, I do research in human subjects psychiatry and I'm not aware of any examples. I'm wondering whether the same could be said in the cancer/oncology field (where many of these tools gained widespread use)?
For extra credit: Did array/seq-related discoveries to the development of any tool that are FDA-approved? Or commercialized? Or used as part of some official treatment protocol or by clinicans at a leading university hospital -- I'm looking for something to legitimize the claim of clinical impact.
Microarray testing for copy number variations is now standard testing for those with undiagnosed intellectual disability. Similarly, copy number testing is becoming standard for various cancers too. However, these aren't transcriptomic tests.
PAM50, as Nathaniel points out, is probably the best example from the array era. But I don't think many diagnostic labs really use it, instead they favor HER2 and ER testing directly.
The Verhaak paper on sub-classification of Glioblastoma Multiforme helped us understand the biology of GBM for sure, but it hasn't really been translated into a clinical test. Current testing focuses directly on EGFR, IDH1 and MGMT.
The problem is that diagnostic pathology is good at giving yes/no answers whereas genome wide studies like array or RNASeq are much more diffuse. EGFR amplification positive by FISH is a much easier sell than "a combination of shrunken centroids and gene set enrichment analysis predict that this GBM is of the proneural(?) subtype." Random forests and SVM just don't fit into current clinical frameworks. What we need are simple, single gene tests that give us a yes or no answer.
Let's bear in mind that discovery to clinic takes several years. RNASeq simply isn't old enough to have made it there yet. A major hope for clinical RNA sequencing is the detection of gene fusions in cancer. For example, we all know the BCR-ABL fusion, but there are hundreds of other gene fusions that predict prognosis, and you can't test for all of them by FISH. My guess is that this type of testing (or a modified form of it) would slip neatly into molecular pathology workflows in the next few years.
The PAM50 array for breast cancer immediately comes to mind. I believe this is or will be in commercial use. I linked the company below.
http://prosigna.com/overview/
Microarray testing for copy number variations is now standard testing for those with undiagnosed intellectual disability. Similarly, copy number testing is becoming standard for various cancers too. However, these aren't transcriptomic tests.
PAM50, as Nathaniel points out, is probably the best example from the array era. But I don't think many diagnostic labs really use it, instead they favor HER2 and ER testing directly.
The Verhaak paper on sub-classification of Glioblastoma Multiforme helped us understand the biology of GBM for sure, but it hasn't really been translated into a clinical test. Current testing focuses directly on EGFR, IDH1 and MGMT.
The problem is that diagnostic pathology is good at giving yes/no answers whereas genome wide studies like array or RNASeq are much more diffuse. EGFR amplification positive by FISH is a much easier sell than "a combination of shrunken centroids and gene set enrichment analysis predict that this GBM is of the proneural(?) subtype." Random forests and SVM just don't fit into current clinical frameworks. What we need are simple, single gene tests that give us a yes or no answer.
Let's bear in mind that discovery to clinic takes several years. RNASeq simply isn't old enough to have made it there yet. A major hope for clinical RNA sequencing is the detection of gene fusions in cancer. For example, we all know the BCR-ABL fusion, but there are hundreds of other gene fusions that predict prognosis, and you can't test for all of them by FISH. My guess is that this type of testing (or a modified form of it) would slip neatly into molecular pathology workflows in the next few years.
The discovery of four molecular subgroups of medulloblastoma (http://www.ncbi.nlm.nih.gov/pubmed/20823417, http://www.ncbi.nlm.nih.gov/pubmed/22134537) are now being considered in clinical trial design (http://www.cancer.gov/clinicaltrials/search/view?cdrid=750566&version=Patient)
This is an important question. There are very limited examples for the success of translational utility, especially in cancer.
The challenge is the dynamic cancer evolution which makes most molecular pathways moving targets. See our recent overall discussions:
Evolutionary mechanism unifies the hallmarks of cancer
Link: http://www.ncbi.nlm.nih.gov/pubmed/24957955
We have specifically studied the mechanism of why it is hard to use specific gene expression profiles to predict cancer evolution. Again, the answer is the unstable cancer genome which differs from a normal genome where the orderly regulation is the key feature.
Unstable genomes elevate transcriptome dynamics
Link: http://www.ncbi.nlm.nih.gov/pubmed/24122714
For more discussion, please see the featured article “Reshaping cancer transcriptome” from GEN. A PDF file can be accessed from our lab page.
http://genetics.wayne.edu/edunew/faculty/heng/lab/News.html
Oncotype DX has FDA approval for breast cancer subclassification and so does MammaPrint. Both come from transcriptomic analysis.
- Badges
- Science topic