When AI Meets Genomic Medicine

With so much hyperbole in the news, it’s difficult to separate false promises from genuine medical breakthroughs. Advances in genomics fall into the latter category.

By John Halamka, M.D., President, Mayo Clinic Platform, and Paul Cerrato, senior research analyst and communications specialist, Mayo Clinic Platform.

The National Institutes of Health describes genomic medicine as “an emerging medical discipline that involves using genomic information about an individual as part of their clinical care (e.g. for diagnostic or therapeutic decision-making) and the health outcomes and policy implications of that clinical use.” Note the operative word “emerging.” Of all the medical specialties that have taken advantage of genomics, oncology is probably at the top of the list, followed by forward thinking pharmacologists who appreciate the value of pharmacogenomic testing to identify gene/drug interactions that can change the therapeutic effects of many medications. Unfortunately, many clinicians in a variety of specialties have yet to see the practical application of genomic medicine.

Between 5-10% of cancers have a strong genetic component and put individuals with specific mutations at high risk. Similarly, patients who have already been diagnosed with a specific malignancy often benefit from gene sequencing. For example, identifying lung cancer patients with an EGFR mutation can help oncologists choose the most appropriate treatment plan for that individual. Such somatic variants are in contrast to germline variants or mutations that put carriers at high risk for specific cancers. Women with specific mutations in the BRCA1 gene, for instance, are more likely to develop breast cancer.

Advances in pharmacogenomics have likewise given clinicians the tools to choose the right drug for the right patient, a topic we have written about in previous blogs. In fact, it’s hard to ignore the preponderance of evidence supporting routine availability of pharmacogenomics data in primary care practice. As a recent Lancet paper pointed out: “Several studies, including randomised controlled trials, have shown that individualising drug therapy on the basis of pharmacogenetic testing leads to improved patient outcomes for specific drug–gene combinations.” 

Machine learning is playing an important role in this emerging field as well. It’s being used to pinpoint genetic variants linked to rare diseases, for example. Gomes et al explain: “…one approach used logistic regression–based machine learning within a large literature-derived data set to match phenotypes to candidate genes in order to help identify potentially causal genes for mendelian disease. Another approach applies maximum likelihood estimation (an iterative method for estimating the parameters of a model) and Bayesian networks (probabilistic graphical models) to achieve the same end.” Studies suggest that ML-based methods may help identify up to 50% of patients with undiagnosed genetic disease.

Mayo Clinic is at the forefront of genomic medicine. Using our current EHR system, providers have access to clinical decision support based on genomic data, which in turn enables them to turn raw genetic lab results into actionable information. Clinicians are also given pharmacogenomic guidelines when they order medication. Equally important, our system supports tumor testing so that oncologists can provide the best possible patient care. 

Of course, we are only at the beginning of the journey. Most actionable genomic data apply to single gene disorders, but most chronic diseases involve a complex soup of genetic risk factors, with no one mutation standing out. That realization has given rise to a variety of polygenic risk scores that may eventually have practical application at the bedside. Equally promising are experiments that have taken advantage of CRISPR, the gene editing tools now being applied to once intractable disorders like sickle cell anemia, hemophilia, and cystic fibrosis.

A recent New York Times article summed up these exciting developments in a few choice words: “Suddenly, It Looks Like We’re in a Golden Age for Medicine.”