Machine learning could leverage genotype and phenotype data to detect disease, drive down false positives from over-testing and bring precision to prescription drugs.
“It’s a boy!” or “It’s a girl!” are pretty common pieces of information people expect to gather in a doctor’s office before their child is born. But what if in the near future those exclamations were followed by information that could impact that unborn child’s lifelong health?
Genomic testing is ushering in that reality, revealing which gene-based issues could arise over a person’s lifetime. This information could follow the soon-to-be newborn through his or her life, determining preventative care measures, identifying effective therapies and shaping their health all the way through to end-of-life care.
This technology is changing fast. As of right now, according to biotech analyst Jonathan Grobert, “Less than 3 percent of the population knows anything about their genomic data. This opportunity is where the internet was in the 1970s.” The Human Genome Project sequenced nearly all of the human genome by 2003, but this effort took 13 years and $1 billion to complete. By 2007, that effort had sped up 70 fold, and it now costs $1,500 and can be completed in about 15 minutes. Soon, according to Dr. Brendan Frey, getting a human genome sequenced will cost less than a trip to the grocery store.
Frey is the CEO of Deep Genomics, a company looking to improve the interpretation not just of genotype — a person’s genetic constitution — but of phenotypes, which range from whether or not you're a redhead to whether or not you have cancer. His company wants to bridge this gap so doctors and patients can interpret and act on genetic information, despite the complex relationship between a person’s genes and their phenotype, including health.
“There’s a huge growth in data sets that allow us to literally peer inside of cells and measure at the single-molecule level what is happening in those cells,” he says. “So there’s this explosion of data that connects genetics to the phenotypes. … The best technology we have for making sense of large data sets is machine learning.”
By applying machine learning to these large data sets, the genomics industry could change the current paradigm of medicine, which is hypothesis driven and tests for only a small number of pathologies while ignoring many others. Frey explains that switching to an information-based model could cut down on unnecessary invasive procedures, since it could do a better job at ruling out false positives.
For example, many women get mammograms. However, the efficacy of getting annual checkups for breast cancer has caused an unintended consequence. The odds a woman has a false positive after 10 years of annual mammograms is as high as 50 to 60 percent, according to the Susan G. Komen Foundation. Through an informatics-based approach, genomic machine learning could better determine which women actually need frequent imaging. Machine learning could parse through a large amount of relevant data to slice down that false positive rate.
An information science-based approach that leverages genotype information could also cut down on prescription rates for drugs. Instead of crudely prescribing the same drug to the masses, genomics could instead drive a precision medicine approach, where doctors understand which patients would most benefit from a drug before writing a script.
In spite of any hurdles, genomics is continuing to get cheaper, quicker and more accessible — even outpacing Moore’s Law. Nine of the 10 leading causes of death, like Alzheimer’s, cancer and strokes, have a genetic factor. Linking this genome information with environmental factors and integrating across the board into the medical field is a big data problem. And it’s one machine learning is equipped to answer, says Frey.
“When it comes to the health care industry, there are practices that will need to change to make that happen. … I think the pressures are there, and the right forces are there to push the community and the industry in the right direction.”
With this rapid pace of innovation in an increasingly crowded market, genomics companies must carve out their niche and differentiate against competitors with a messaging and positioning that sets them apart. In this rapid race for industry mindshare, genomics companies have a once-in-a-lifetime opportunity to make their mark. It makes for an exciting time for genomics companies to step out and become a thought leader in the field, contributing to the industry dialogue and shaping the market for generations to come.
Follow Alisa Valudes Whyte on Twitter: www.twitter.com/MerrittGroup