HUNTSVILLE, Alabama (GenomeWeb) – There's a continuum of phenotypes that stretch from Mendelian disease to related conditions that may have clinical implications, according to work presented yesterday at the Genomic Medicine Conference being held by HudsonAlpha here.
Vanderbilt University's Nancy Cox developed a gene by medical phenome catalog by applying the PrediXcan analytical approach she and her colleagues previously published to 10 years to 15 years' worth of electronic health record data from some 215,000 people. This enabled them to predict the expression of certain genes in 44 different tissues to predict other conditions that might later arise in patients as well as identify potential therapeutic approaches. They also noted that the reduced expression of genes typically linked to Mendelian disease — where gene expression is usually lost — is also associated with similar phenotypes that they can glean from their analyses.
"It really changes the way you think about medicine," Cox said of the approach.
PrediXcan, which Cox and her colleagues described in Nature Genetics about a year ago, is a computational method that couples gene expression and phenotype data to predict which genes might play a role in a heritable condition. In particular, Cox said it estimates the part of gene expression that is genetically regulated, rather than affected by the environment, and correlates that to the trait of interest. It then prioritizes genes that are likely to be involved.
As she noted at the meeting, Cox has applied this approach to Vanderbilt's DNA database, BioVU, which houses more than 215,000 samples. The database also contains some 10 years to 15 years' worth of electronic health record data, including genotyping as well as whole-exome and whole-genome sequencing data.
Together, this yields a comprehensive gene by medical phenome catalog, Cox said. They tested the expression of genes in 44 GTEx tissues for associations with medical phenomes.
In one example she described, her analysis enabled her to make predictions about follow-on symptoms or conditions patients could develop. Variants in NFIX have been linked to Marshall-Smith Syndrome and Sotos syndrome, and both conditions are marked by certain bone features, muscle weakness, distinctive facial features, breathing difficulties, and intellectual disability or delay. Cox found that NFIX is highly expressed in muscle, heart, and brain, as well as in the cervix, uterus, and fallopian tubes.
Based on this, she predicted that there might be cardiac and esophageal involvement, and that in girls uterine and pelvic inflammatory disease may also develop. This, she said, could be useful information to know as the patients age.
In another example, Cox examined SLC39A4, which is associated with autosomal recessive acrodermatitis enteropathica, which is marked by rashes near bodily openings. SLC39A4, she noted, is a known zinc transporter and acrodermatitis enteropathica can be treated with zinc.
But in some people, SLC39A4 expression is decreased, though not entirely lost. In this case, Cox's analyses linked it to anemia, schizophrenia, and known consequences of zinc deficiency like cardiomyopathy and other blistering skin conditions such as impetigo and cellulitis, but also to type 2 diabetes and diabetic neuropathy, among conditions.
An innocuous therapy — zinc — could treat this, she said.
Acrodermatitis, she added, affects about one in 500,000 people and there are none in BioVU, but there are some in the database with reduced gene expression.
Cox said she and her colleagues had been considering launching a trial to test zinc as a treatment for people with reduced SLC39A4 expression and phenotypes echoing that of the Mendelian disease.
But, she noted that they might instead incorporate it into Vanderbilt's Pharmacogenomic Resource for Enhanced Decisions in Care and Treatment (PREDICT) program. That project is genotyping patients for certain variants, and when their physician goes to prescribe them a drug, the electronic medical record system alerts the physician if they have any genotype with implications for drug response.
So far, she noted, doctors follow this guidance about 70 percent of the time, and when they don't they've found that patents tend to have worse outcomes.
They are considering adding SLC39A4 predictions to this program.