The following blog article was written live at the 2nd Annual Symposium of RNA in Precision Medicine at the University of Michigan through the Center for RNA Biomedicine. Check out MiSciWriter blog for the whole coverage of the symposium and follow the #umichRNA on Twitter. 

“Integrating Genome, Transcriptome and Electronic Health Records for Discovery and Translation.” Nancy Cox, PhD, Vanderbilt University.

Blogger: Sarah Kearns
Editor: Molly Kozminsky

Nancy Cox is a professor of medicine and director of the Division of Genetic Medicine in the Department of Medicine at Vanderbilt University. She was recruited to Vanderbilt in 2015 to lead the Vanderbilt Genetics Institute (VGI). The VGI’s mission is to “promote genomic discovery and advance understanding of the human genome” and as such, Dr. Cox develops of methods to better analyze genomic data. 

The human genome is becoming easier and cheaper to obtain. However, the ability to predict which genetic variants or small mutations will cause a significant change in observable characteristics, or phenotypes, has been a very difficult task. This is partially because there is a lot of redundancy in our protein sequences to have a buffer system in place to make sure small transcriptions mistakes do not severely harm us. For example, a single nucleotide mutation will not necessarily result in an incorrect protein product.

Dr. Cox studies the interaction between the genome, the full DNA sequence, and the transcriptome, the gene products that are actually formed. This is difficult to study, especially in humans, because: 1) each individual has a unique DNA sequence, and 2) depending on modifications, it may or may not be possible for a specific gene to be transcribed.

Mapping DNA and transcription products can be done with genome-wide association studies (GWAS), which is an examination of all the genetic variants of different individuals across an entire genome to see if variants are associated with a trait. For an example, a study might focus in on a certain DNA mutation and see if it correlates to a phenotype, or an observable characteristic, in a person. The Cox lab in particular has developed a method called PrediXcan that is able to predict the amount of gene expression based on an individual profile.

Many environmental exposures can affect our epigenetics. Factors such as smoking can affect an individual’s DNA without actually changing the genome. However, using the open source software PrediXcan and GTEx, we can get a good sense of a prediction of phenotype from gene expression.

A Phenome-wide Association Study (PheWAS) takes a single gene mutation product and scans against all of the medical records that contain that gene, essentially annotating the medical genome. Researchers are trying to probe the biology of the gene, not only on an individual, but trying to scale that up to be able to predict disease state phenotypes across a population. With the databank at VGI containing 2.6 million medical records containing everything from images to pharmaceutical records, they are certainly able to get a big picture of the medical genome.

With this information they are able to map a phenotype, for example acidosis, to a specific gene. This was validated by taking a gene and checking downstream expression and subsequent predicted phenotype. However this also suggests that there are multiple gene products that lead to the same phenotype which opens the door to personalized or combinatory medicine.

Knockout studies are a technique to study a gene by effectively deleting it to see what happens. Dr. Cox’s did a knockout study with zebrafish and silenced the gene encoding for eyes.  Eyes in particular have many different phenotypes and as such, this knockout could have many different effects. Afterwards, the researchers tracked to see what, if any, other effects were seen within the fish. Most notably and surprisingly, they saw a few different phenotypes. A fraction of fish ended up with only one eye, some fish had one normal eye and one small, and other fish had no eyes at all.

We’re starting to see that DNA inheritance can fall on a spectrum from classic Mendelian genetics to something much more complex. Consequently, a knockout study can yield a range of phenotypes including losses of function, deleterious function, and simply lower gene expression.

Cox went onto study these genomic overlaps in human conditions such as skin blistering and schizophrenia and was able to see associations to other similar phenotypes. This means that a person with a Mendelian disease might have features that become contributing factors to other similar biological problems. However, there is a bigger population that, instead of having an easy dominant-recessive (yes-no) phenotype for a gene, just has lower gene expression. This could mean that these patients, instead of needing pharmaceutically derived drugs, could use dietary supplements to help their body return to a non-diseased state.

A PREDICT-type Trial was done to identify, through phenome patterns, the patients who are most likely to have a reduced gene expression. Cox was able to apply this to neuropsychological phenotypes focusing on fear and phobia. She found an interesting overlap between genes related to phobia with those related to heart problems and severe allergies. Unfortunately, there was an abundance of heart problems that were potentially misdiagnosed as phobia and panic attacks, especially in young women.

With PREDICT being a personalized way to measure RNA interaction with the genome, we are able to focus in on what specific gene expression within an individual that can lead to coordinate regulations. When this data is collected, scaled up, and applied to a large biobank, the VGI is able to have an integrated data set to both raise awareness for these comorbidities as well as with discovering better medicinal treatments that are able to cover the unique pattern of expression of an individual.

Sarah Kearns is a first year in the Chemical Biology Doctoral Program at the University of Michigan. Currently, she is doing research rotations to find a long-term lab environment ideally focusing on enzyme structure-function relationships for drug development applications. You can find her on Twitter (@annotated_sci), LinkedIn, or at her website Annotated Science.