Building Human Longevity’s Exhaustive Knowledgebase
By Kenneth Buck
June 17, 2016 | Despite advances in genomics in the past decade, there are still large portions of the genome that scientists and physicians don’t know how to interpret. Human Longevity Inc. (HLI) wants to fill in those gaps. In honor of the company’s progress, HLI was awarded the Best Practices Judges' Prize for its HLI Knowledgebase and Health Nucleus platform for the development and delivery of individualized medicine at April's Bio-IT World Conference.
“It’s nice to have the independent recognition of the solution that’s been developed in the last two years,” said Yaron Turpaz, CIO of HLI, who also participated in the keynote panel on precision medicine. “What we are trying to do is to truly transform healthcare in many different aspects and move it into preventative healthcare, and some of the informatics solutions that we are developing are enabling such approaches.”
Realizing these ambitions will require interpreting each patient's profile in relation to petabytes of medical knowledge from large populations. And so HLI is building a rich repository of multidimensional genomic-phenotypic associations by bringing together next-generation sequencing, alignment and knowledge management tools, and cloud-based storage, search and analytics infrastructure. HLI recently completed a $220 million offering of Series B Preferred Stock, and it plans to grow its commercial and research services and adds staff.
HLI is committed to sequencing 1 million genomes by 2020 and integrating standardized, longitudinal phenotypic records. The Knowledgebase currently contains approximately 25,000 genomes and 11,000 integrated health records (IHR’s). At today’s whole genome sequencing speeds, intake of IHR data is the rate-limiting step. But the company is working on accelerating both streams.
Building the IHR
Each patient who visits the Health Nucleus clinical center in San Diego, California, meets with physicians and genetic counselors trained in bioinformatics, and undergoes whole genome sequencing and analysis of their metabolome, microbiome, and lab diagnostics including a full-body MRI. Turpaz indicated the potential for additional Health Nucleus centers around the globe.
Another component of HLI’s business plan is to offer pharmaceutical companies and other entities access to the growing Knowledgebase in return for shared data. HLI’s partnerships with AstraZeneca (to sequence and analyze up to half a million samples collected in clinical trials), Genentech (to sequence genomes of clinical trial patients), and Discovery Insurance in South Africa and the U.K. will augment the breadth and depth of clinically relevant source data and populations represented in the Knowledgebase. Both the clinical informatics and the personalized medicine aspects of HLI are being administered under research protocols.
Meet the Medical Avatar
The Knowledgebase has two “entry points”—one for scientists in pharmaceutical and biotech companies, and a second for patients and their physicians. Cloud implementation on Amazon Web Services (AWS) allows distributed real-time search and visualization. Patients and physicians see representations of their genome and phenotypic profile through Health Nucleus’ Medical Avatar application now available in the iTunes store.
“[It’s] a truly personal, photographic, and technology-driven representation of medical results,” and has been evolving based on feedback provided by experts, said Turpaz. The Medical Avatar dynamically updates with changes to patients’ records and growth in the HLI Knowledgebase.
Turpaz noted Medical Avatar’s role in helping patients and their primary care doctors to understand the information they receive.
“The platform we’re building should help educate both the patients and the physicians, but should be as self-explanatory as possible,” he said. “At the same time, it’s essential for the physician to understand the background and what stands behind integrating this information. There is an education process, and that’s where our HLI-trained physicians come into play,” to facilitate joint discussions.
Risk analyses are performed using data in the Knowledgebase and literature mining, and are presented to the physician graphically with confidence levels and links to relevant publications. “Initially, each component is being analyzed independently and then we provide an integrated analysis,” said Turpaz. Kinship information is incorporated into the Knowledgebase as patients often recruit family members to enroll, which along with cohorts such as the TwinsUK study, will allow inherited risk analysis. When available, data from environmental exposure and epigenetics studies are processed, although costs currently prevent comprehensive epigenetics collection.
For cancer patients, HLI works with their oncologists and provides whole-genome analysis of the tumor and the germline, as well as RNA-seq and immunoSEQ (an Adaptive Biosciences kit for sequencing the adaptive immune system), to predict the patient’s response to immunotherapy and/or to match the tumor’s genomics to available drugs. HLI is developing automated integration approaches to speed this process, and will help patients locate clinical trials that are recruiting subjects with specific mutations.
On the Horizon
As the Knowledgebase grows, so does the potential for accelerating clinical trials. The Knowledgebase is designed to enable biomarker discovery, target identification and in-silico validation. Access to multidimensional variant associations in a million-genome database will permit researchers to select specific cohorts for analysis with strong statistical power.
“You have the inclusion and exclusion criteria in the clinical study, and those can be related to… diagnostics that can guide you in what you’re analyzing, how you’re analyzing,” said Turpaz. “In different stages of the trial, different elements of this knowledge can be utilized. So with time, we can work with statisticians who are designing clinical trials and help define endpoints.”
HLI is also developing machine learning approaches, whose performance is expected to improve as the Knowledgebase balloons. After being trained on linked datasets of 3D facial scans and genomes, machine learning demonstrated the ability to predict physical facial characteristics of a patient from their genome. But predicting changing disease risks as a function of aging are the goal. “We have some initial discoveries, some will be published soon, and yet we know it’s just the beginning,” hinted Turpaz.
Turpaz pointed to rare diseases as another potential area to exercise the Knowledgebase. “We started to [look at] some specific elements of undiagnosed rare diseases to better understand the genomics in infants. And this is something we’re building a platform around, but it hasn’t been formally launched. We’re in discussion with different children’s hospitals and trying to understand this better now.”
