NEW YORK (GenomeWeb) – GenomeSmart, a Los Altos, California-based company founded by husband and wife team Vandana and Sanjay Sathé, is looking to partner with stakeholders in the healthcare industry interested in using its digital risk assessment technology to ensure patients are getting guidelines-supported information about their genetic testing options.
The startup, which has been in operation for more than a year, has developed an artificial intelligence-based platform, called GenomeBrain, which digitizes the pretest process. Typically, a genetic counselor or a healthcare professional gathers information from patients during that process about their personal and family history of certain illnesses. Based on this information, the counselor determines if patients need testing to diagnose a genetic condition or assess their predisposition for a disease, and which tests are most appropriate.
Earlier this month, the company launched the GenomeBrain platform, which was developed with input from genetic counselors and is designed to mimic what they do during the pretest process. GenomeSmart CEO Sanjay Sathé said he thinks this platform can be useful for healthcare systems, labs, payors, and large employers, which all have been impacted by the increasing availability and utilization of genetic tests.
According to CSO Vandana Sathé, the risk assessment tool takes individuals through a seven-minute evaluation during which it asks them a variety of questions to collect information, such as their ethnicity, family history of illness, age, and if they’ve had prior genetic testing. The machine-learning algorithm then maps this information to existing guidelines, for example from the National Comprehensive Cancer Network (NCCN) and the American College of Obstetricians and Gynecologists (ACOG), and determines first if the individual should be tested at all.
For those who should be, the GenomeBrain platform provides test options and cites specific guidelines that support them. The platform doesn't recommend any brands of tests but identifies specific genes or panels of genes. If GenomeBrain is being deployed within a healthcare system, for example, the test recommendations are sent to patients, as well as their physicians or genetic counselors who order their testing.
"In a typical seven-minute assessment, the machine-learning algorithm goes through millions of variables, including mapping the user profile to the national medical guidelines and all available types of tests, to produce a personalized report in milliseconds with the test recommendation," said Sanjay Sathé.
The GenomeBrain platform is currently available for identifying test options for hereditary cancer risk and tests in the context of reproductive genetics. The company plans to expand the platform into heart conditions and pharmacogenomics.
To develop the platform, the firm brought in seven genetic counselors who work within healthcare systems to learn their processes for evaluating patients and to formulate the questions the platform would ask patients during the seven-minute assessment. The genetic counselors went through a couple of hundred use cases where they picked tests based on guidelines for a patient profile and trained the machine-learning algorithm to do the same. "Going through those datasets, we identified conditions that were to be met to be flagged for being at high risk," Sanjay Sathé said.
The company further refined the model through an iterative process by splitting the genetic counselors into two teams. The first group was given use cases and their test recommendations were compared to the tests the algorithm picked, which allowed the company to calibrate the algorithm "by reducing the delta between ‘what was recommended' and ‘what should be recommended,'" Sanjay Sathé said. The firm then repeated this with the second set of genetic counselors.
Based on this experience, the platform's test recommendations aligned with genetic counselors' test picks 97 percent of the time. Sanjay Sathé attributed "the very slight difference" between counselors and the algorithm to the difference of opinions between genetic counselors rather than the algorithm. "As with all AI or machine-learning algorithms across industries, the algorithm will continuously be learning and refining itself over time," he said. The company hasn't yet published on the platform's validation but plans to do so after amassing more data.
Amy Curry Sturm, president of the National Society of Genetic Counselors (NSGC), said she has discussed the technology with GenomeSmart's founders but hasn't yet seen a demo of the latest version. She would like more data on the technology, for example, how the platform does when dealing with straightforward and complex cases.
"Some situations are really nuanced and it's not always a straightforward answer" as to which tests patients should get, she said. Sturm, who is also co-director of the MyCode genomic screening and counseling program at Geisinger Health, suggested that GenomeSmart validate its platform prospectively on real patients with a variety of clinical scenarios.
Sturm has experience deploying technological solutions within the MyCode precision health research program at Geisinger, which serves largely rural communities in Pennsylvania and New Jersey. The health system partnered with San Francisco-based Clear Genetics to develop a chatbot to interact with patients who enroll in its research program, get their exomes sequenced, and learn about certain actionable results. The Genetic Information Assistant, or Gia, serves as a technological aid to Geisinger's team of more than 25 genetic counselors, whose responsibilities include interpreting variants, delivering results to participants, and speaking to at-risk relatives.
The health system currently uses Gia to check in with MyCode participants one month after they've received their test results to gauge what actions they've taken. Additionally, patients with positive results can use the bot to relay pertinent educational information to at-risk relatives. Sturm indicated that her group will soon publish data on the deployment of Gia in these settings and present data at the NSGC annual meeting in the fall.
Geisinger will likely deploy the chatbot next as a tool for consenting participants for the MyCode study, but before doing that, it will compare Gia's ability to consent participants head-to-head against precision health associates who do this work. The health system is also working with Clear Genetics to potentially use Gia to identify, ahead of a genetic counseling session, the most important concerns or questions patients want to discuss with counselors.
GenomeBrain can also be used to optimize genetic counselors' time or fill a need for genomics expertise in regions lacking it, said Shannon Kieran, executive VP at GenomeSmart. At some institutions, patients have to wait for months to speak to a genetic counselor, for example, and the technology platform can be used in these cases to determine quickly if patients have a clear indication for genetic testing and if they need further evaluation, she said.
"The assessment for testing is a huge educational feat that genetic counselors undertake daily in a very redundant way," Kieran said.
Kieran, herself a board-certified genetic counselor, isn't concerned that the GenomeBrain platform will replace her professional colleagues, but she believes it could be a valuable aid to them and other time-strapped medical professionals. "This is a tool that I think will enable genetic counselors, and other healthcare professionals who may or may not be up to date on the latest in the genetics space, to really practice at the top of their field."
According to one estimate, there are 74,000 commercially available genetic tests in the US and 14 new tests are launched on the market daily. Meanwhile, there are more than 4,000 genetic counselors in the US and Canada. Although the US Bureau of Labor Statistics estimates that jobs in this sector will increase by 29 percent between 2016 and 2026, industry projections suggest that the genetic testing field will also expand during this period, creating demand for genomics expertise and ancillary services.
"There is always going to be complex cases. There's always going to be the need for human assessment and analysis for some," said Kieran. "But we have an issue that we need to solve in terms of scaling genetic services in the healthcare industry."
As genetics becomes further integrated into patient care, this need for expertise, education, and support services is increasingly being felt by stakeholders across the healthcare industry, and many are turning to telemedicine and AI-powered technologies.
For example, a number of consumer-facing genetic testing companies now allow individuals to initiate orders for health-related genetic tests online. But those consumer-initiated orders are then reviewed by doctors at telehealth services to ensure the right test is being performed. Customers can also speak to genetic counselors over the phone if they have questions ahead of testing or if they want to discuss test results.
Optra Health, based in San Jose, California, earlier this year demonstrated an AI-based clinical report interpretation and guidance system, called GeneFax, that allows consumers and providers to query genetic data via Amazon's Alexa and Microsoft's Cortana. DNAFeed is another company that provides genetic counseling and pharmacist support services via phone, video, and online chat, but in the chat platform, an AI assists the human experts by coming up with some initial answers to patients' questions that they can review, edit, and use.
GenomeSmart has also developed an educational platform, called GenoLearn. Patients who are assessed by GenomeBrain can use GenoLearn to get educated about basic genetics concepts as well as hereditary cancer and reproductive genetics topics.
Vandana and Sanjay Sathé founded the company after recognizing that although the genetic testing industry was growing at a rapid clip, consumers lacked in their understanding of genetics and had little knowledge of the strengths and limitations of different kinds of tests. Especially after two of Vandana Sathé's close friends were diagnosed with cancer, it hit home for her that they didn't have an objective resource for determining if they should get genetic testing and which tests should be performed.
"We've had conversations with couples where they suddenly had a problem during pregnancy and they didn't know what to do," added Sanjay Sathé. "There was no comprehensive resource apart from some brochures they got at the clinic."
GenomeSmart's founders plan to market their technology platform to healthcare stakeholders impacted by the explosion of genomics knowledge. For example, GenomeBrain could potentially address the needs of healthcare systems interested in driving operational efficiencies, payors that want to ensure tests they deem to be "medically necessary" are ordered, labs that want to identify the most appropriate tests in their portfolio of products for patients, and self-insured employers that want to provide employees services that keep them healthy and reduce their downstream financial exposure.
The company has not yet announced any partnerships yet, however. GenomeSmart currently has seed funding from angel investors and is about to initiate another seed funding round. Once the company signs contracts with customers, it will head into a series A round, according to Sanjay Sathé.
"We've seen interest from all four groups," he said. "There's an opportunity to help a lot of people."