Shiva Amiri, PhD, Director Research Infrastructure, 23andMe
As our ability to generate and collect data from multiple sources has increased rapidly over the past three decades, this has presented new opportunities to use that data to benefit our health, environment, productivity, and most facets of life today. This has also resulted in the rapid growth of technologies that enable the management and analytics of these large datasets including cloud infrastructure and novel machine learning (ML) methods and technologies.
This is an exciting but also complex time for every industry, company, and consumer. Exciting because new technologies, together with the data, enable new capabilities and insights which could be powerful. For example, in improving sales, or optimizing our favorite routes on ride sharing apps, and even things like identifying abuse and fake content online. But also complex, because understanding the technologies, knowing how to deploy and use them effectively can be a challenge. In addition to the often grey ethical and policy landscape that can be difficult to navigate.
The retail space has been quick to leverage these technologies as ML has made clear financial impact and the policy and ethical landscape is less complicated than many other sectors. Two of the biggest players and technical leaders in the sector are Amazon and Netflix, both generate custom recommendations (recommender models) for their users based on existing buying patterns and other data including demographics.
The investments the retail sector has made in ML have resulted in rapid technical advances; however the use of machine learning has not been limited to the retail space. We see the application of machine learning in most sectors today, including biotech, health, aerospace, mining, and media.
As someone on the biology and health side of data science, I’ve witnessed significant growth in the number of ‘data companies’ over the years. Companies and research organizations are recognizing the value of data and related technologies such as machine learning for research, discovery, and health outcomes - all of which can derive significant scientific and financial value.
With the right balance of partnerships with technical leaders and growing in-house capabilities, organizations can benefit from the deployment and use of machine learning to enable new discoveries
Today, as we see the traditional tech “giants” -- Apple, Amazon, Google, NVIDIA and more -- leaning into health and biology in big ways they can bring their large-scale data and data science platforms to these application areas. As a result, there’s a smaller divide between the technology and health industries. For instance, Amazon’s recent collaboration with the Fred Hutchinson Cancer Research Center is intended to evaluate “millions of clinical notes to extract and index medical conditions.” In fact, most of these companies could now also be classified as biotech and health tech firms in some capacity.
Personal devices that capture health related data at growing rates -- such as sleep, electrocardiography (ECG), behavioral data, steps and more -- also present valuable data types that companies and research institutes are using to enable insights and discoveries in health and precision medicine.
Diagnostics continues to evolve with the advancement of technologies like machine learning. Specifically, the rapid growth of deep learning methods (a method of machine learning) in imaging diagnostics for decision support at clinical sites in addition to deployments for remote health applications as a first line of diagnosis - examples include Stanford Artificial Intelligence Laboratory’s skin-cancer diagnostic product published in Jan 2017 issue of Nature.
Genetic information from customers who consent to participate in research is increasingly available and accessible. The combination of genomics and electronic health records/ medical history data has benefits in the health and therapeutics space. This data, layered with machine learning techniques, enables new powerful insights where we begin to predict predisposition to disorders, optimize cohorts for clinical trials, understand medication response, design drugs and more. It’s an exciting time to be a part of this field.
One of the main reasons I joined 23andMe was because of its focus on research and therapeutics development, made possible by customers who consent to share their genetic and phenotypic data for health discoveries, and enabled by the unique research and compute platform at 23andMe. This large-scale research infrastructure enables the data processing and computations (including building and deploying models) for our consumer facing products but also for our research and therapeutics teams.
23andMe’s consumer business provides health and ancestry reports to customers. Some of the reports predict predisposition to conditions, including the recent Type 2 Diabetes Health Predisposition Report, while lighter ‘trait’ reports could tell you if you’re likely to get motion sickness or match musical notes. As more customers consent to participate in research, these models can become increasingly sophisticated.
The impact to date and the potential of machine learning in biology, health and therapeutics is vast, as it is for many other sectors including retail, manufacturing, and media. With the right balance of partnerships with technical leaders and growing in-house capabilities, organizations can benefit from the deployment and use of machine learning to enable new discoveries to benefit humanity, gain valuable insights for businesses, and generally allow predictive analytics to play a bigger role in their organizations.