Healthcare’s AI revolution is changing the way we think about age-related diseases, even aging itself
We are in the midst of an epidemic. Regardless of your family history, race, or geography, there is a disease that will befall each and every one of us. You can hide in the mountains of Siberia, but the disease will still reach you because it’s not contagious. It’s followed humanity throughout time, and will continue to do so into the foreseeable future despite our recent attempts to forestall it.
That disease is called aging.
It’s counter-intuitive to consider aging a disease because we equate it with time’s inevitable onslaught, but there is a growing contingent within the scientific community who believes the two concepts are separable. In fact, so separable that age may one day truly just be a number. As reported by The Guardian, the best forecasted life expectancy among 35 developed countries is over 90 years old among women born after 2030, an increase of 6.6 years over those born in 2010. [1]
Indeed, according to James Vaupel, director of the Max Planck Institute for Demographic Research and professor at Duke University, “for 160 years, best-performance life expectancy has steadily increased by a quarter of a year.” [2] It might be argued, and it has, that we are steadily reaching a maximum age range; however, Vaupel retorts, “if life expectancy were close to a maximum, then the increase in the record expectation of life should be slowing. It is not.” [3]
The idea that humanity can solve aging sounds absurdly arrogant. Overlooking the existential question of the determinant of life’s meaning without a definitive end, death has been a looming certainty universal to the human experience — the only certainty.
Photo by Andres Urena on Unsplash
Regardless of its apparent absurdity, the largest tech companies are entering the race to end aging. Google’s Calico Labs was launched in 2013 as a moonshot venture intent on addressing age-related diseases and extending human life. But Google isn’t alone. Apple has made consumer healthcare an area of intense focus as well. Tim Cook, Apple’s CEO, stated in an interview with CNBC,
“if you zoom out into the future, and you look back, and you ask the question, ‘What was Apple’s greatest contribution to mankind, it will be about health.” [4]
Indeed, it would appear that the field of medicine is becoming a treasure trove of quantifiable data. It’s not just health records, family history and bloodwork that are of interest to researchers, but the very code of life itself — DNA. To paraphrase Siddhartha Mukherjee, Pulitzer Prize-winning author and biologist: how appropriate that the study of the smallest bits of life, our genes, merge with the field of computer science, a field revolutionized by the understanding and manipulations of bits comprised of 1’s and 0’s.
As we’ve come to understand about our personal information, data is invaluable, particularly because it’s information from which deeper knowledge can be gleaned. If multi-billion-dollar industries have arisen from buying and selling our browsing data on the internet, how much more valuable must be the secrets hidden in the data of our genome? To be clear, I don’t speak here of the monetary value of an individual genome (however concerning that idea may be), but of the wealth of insight that can be gained by mass statistical analysis of millions of genomes and their accompanying physical expressions.
Photo by Taylor Vick on Unsplash
The move toward a data-centric approach in every industry was bound to edge its way into the medical field. According to Time Magazine in its piece, “Google vs. Death”, “medicine is well on its way to becoming an information science”. [5] And with all that information, it’s only natural that an organization like Google, which views “information as a kind of commodity, a utilitarian resource that can be mined and processed with industrial efficiency,” [6] would position itself to make use of all that precious data.
Although Nicholas Carr, a prominent writer on technology’s impact on modern society, speaks negatively about the consequences of outsourcing humanity’s thinking in his piece, “Is Google Making Us Stupid?”, he notes precisely why the realm of medicine would be of interest to Google. Being “a fundamentally scientific enterprise, Google is motivated by a desire to use technology, in Eric Schmidt’s words, ‘to solve problems that have never been solved before’.” [7] Among the most difficult problems to solve is aging, or at least age-related diseases, and many medical professionals believe the solutions lie at the intersection of artificial intelligence and genomics. Carr himself quotes Google’s co-founder, Larry Page, in a speech where he states that “working on search is a way to work on artificial intelligence”.
Solving for AI is not a goal to be achieved for the sake of its own fulfillment, but for its application to broader, wider-ranging problems facing humanity that humans alone do not have the capacity to solve.
Artificial intelligence is shifting the landscape of what Carr dubs “knowledge work” across all industries and healthcare is no exception. However, in his response to Carr, notable futurist, Jamais Cascio, paints this shift as a necessary evolution in “how we manage and adapt to the immense amount of knowledge we’ve created”. [8] In the case of the genome, the knowledge created is the massive information stored across millions of sequences and their corresponding statistical significances — the very information that we ourselves are comprised of.
Photo by Amelie Ohlrogge on Unsplash
Cascio goes on to make a prescient statement, telling us that “humans won’t be taken out of the loop — in fact, many, many more humans will have the capacity to do something that was once limited to a hermetic priesthood”. This has proven to be true in the context of medical research. Derek Lowe, a drug discovery researcher, tells the New York Times,
“It is not that machines are going to replace chemists, it’s that the chemists who use machines will replace those that don’t”. [9]
Technology may not grant us immortality, but the increasing interdependence of human and machine intelligence will usher us into a new world of rapid medical advancements — a future with more affordable, personalized, and safer preventative care where both our lifespans and healthspans are drastically increased.
Medical researchers have long sought to decode the human genome. The first full human genome was sequenced in 2003 which took 10 years and cost $3 billion. As Anshul Kundaje, assistant professor of genetics and of computer science at Stanford, points out in his 2016 talk at “The Future of Artificial Intelligence”, “today we can sequence genomes in a matter of a few days for less than $2,000”. [10]
The rapid reduction in cost has created an explosion of genomic data — data upon which machine learning algorithms can gain insight. Such insights have the potential to save patients’ lives, as their genetic information can assist practitioners in prescribing the most effective drug to an individual while minimizing side effects. However, simply decoding the genome isn’t enough to advance medicine in the ways many hope; researchers must understand how the snippets of code are expressed in order to utilize the sequence of 20,000 genes meaningfully.
Photo by Daniel Christie on Unsplash
Kundaje explains that his team took the raw data, applied machine learning models and was able to assign “some kind of integrative functional annotation of the genome”. By doing so, his team identified potential targets to address Alzheimer’s — a hereditable disease that is the target of many gene therapy efforts and whose mechanism of action has long escaped traditional treatments.
Knowing where to look and predicting how a drug will behave provides pharmaceutical researchers an enormous advantage when developing effective therapies. The ability to directly edit or turn off snippets of the genome associated with diseases has caused much excitement in the field of genomics. A system called CRISPR-Cas9 has received a lot of attention “because it is faster, cheaper, more accurate, and more efficient than other existing genome editing methods”. [11] However, the ability to precisely edit genes is useless to researchers without a target. Projects like Kundaje’s can help acquire these genetic targets.
Although DNA is at the kernel, it’s not the entire story behind disease. DNA directs proteins (utilizing RNA as an intermediary to transcribe and translate instructions), and proteins perform as directed. But, in the process, “they can become tangled…leading to disorders such as diabetes, Parkinson’s and Alzheimer’s disease”. [12] For this reason, predicting the shape of a protein is vitally important to understanding its function — a task historically of extreme difficulty.
In 2011, researchers crowdsourced the matter, turning to gamers to help find solutions to protein folding problems. Within three weeks, the combined efforts of gamers and researchers resulted in “crucial insights to solve the structure of a protein-sniping enzyme critical for reproduction of the AIDS virus” — a feat that had stumped scientists for a decade. [13] According to Scientific American at the time, “humans retain an edge over computers when complex problems require intuition and leaps of insight rather than brute calculation”.
Protein Folding from Wikimedia Commons
Apparently, that edge didn’t last the remainder of the decade. Google’s AlphaFold, a deep learning algorithm trained on the aforementioned Foldit game, beat out 98 competitors in a protein-folding challenge by an enormous margin. “Predicting the most accurate structure for 25 out of 43 proteins, compared with three out of 43 for the second placed team”. [14] Perhaps more importantly, the program went from taking weeks to supply a prediction to doing so in a matter of hours. With more data and more practice, AI has proven it will gain accuracy and efficiency.
Insight into protein folding allows researchers to peer not only into the function of hereditable diseases, but also how to develop drugs to treat them. As reported by the New York Times, “If scientists can predict a protein’s shape, they can better determine how other molecules will ‘bind’ to it”. Made more widely available, tools like AlphaFold can drastically decrease the amount of time and money required to develop treatments for previously incurable diseases.
While AlphaFold’s dramatic win signals AI’s increasing importance in biochemical research, not all are convinced that machines and those developing them will be able to provide all of the answers. As Dr. David Baker, director of the Institute for Protein Design at the University of Washington, expresses to the New York Times: “creating proteins is more important to drug discovery than the ‘folding’ methods being explored,” a task he believes AlphaFold is not well-suited to complete.
As witnessed with earlier doubts regarding the problem-solving ability of AI, time will tell.
IBM Watson from Wikimedia Commons
Machine learning and its application in medicine is not limited to computational genomics or protein folding. In 2016, Watson, IBM’s artificial intelligence machine, did what doctors couldn’t. Within 10 minutes, Watson analyzed 20 million cancer-related research papers and delivered a correct diagnosis — a feat that doctors of the University of Tokyo had failed to accomplish in the several months of treating their 60-year old patient. [15] Similar to AlphaFold, Watson also increased in accuracy and efficiency, with claims by IBM of 240% improvement in performance between 2011 and 2013.
As we’ve seen between the examples of Watson and Google’s DeepMind, these AI systems perform better with larger amounts of data to analyze and compare. In addition to developing these deep learning algorithms, it’s equally important to collect and make available as much health data as possible.
According to the National Institute on Aging, the larger the pool of genetic information from both healthy and at-risk patients, “the more clues [researchers] will have for finding additional risk-factor genes”. Kundaje also mentions in his talk that machine learning algorithms will have to be fed both personal and large-scale genomic data as well as health records to be effective in recognizing statistically significant correlations in the bits of our genetic makeup and the diseases that errors in those bits can cause.
Thankfully, institutions are working together to make this data easily accessible to researchers. The ENCODE (Encyclopedia of DNA Elements) Project is a public research consortium whose research database is freely accessible. Meanwhile, Harvard launched the Personal Genome Project in 2005, a voluntary program to which participants provide their biological samples for genetic research. Hopefully, the wide availability and growing database of health information, along with integrated digitization of health records, will fuel the advancements of AI such as IBM’s Watson and Google’s DeepMind.
Armed with this data, machine learning will be a key partner to both doctors and drug researchers in, as Kundaje claims, “transforming this data into essentially predictive prognostics, diagnostics, smaller drug targets, and optimal treatment strategies”.
Although not yet beyond its technical hurdles, AI will usher us into a new era of medical advancements, making individual treatment more personalized and effective, and fundamentally disrupting the patient-provider relationship, not by replacing medical providers, but by augmenting their expertise and allowing them to treat patients more effectively. With the emergence of AI in the healthcare sector, researchers will be empowered to analyze data that they never could before, tapping into the very code of our being — DNA — to understand what makes us healthy and whole, eventually making personalized medical treatments a reality with the potential to address the most ubiquitous disease of all — that of aging.
This article originally appeared in Towards Data Science
References
[1] B. Sarah, “Life Expectancy Forecast to Exceed 90 Years in Coming Decades”, The Guardian (2017), www.theguardian.com/society/2017/feb/21/south-korean-womens-life-expectancy-exceed-90-years-2020-study
[2] J. Oeppen and J. Vaupel, “DEMOGRAPHY: Enhanced: Broken Limits to Life Expectancy” (2002), Science, vol. 296, no. 5570, pp. 1029–1031, https://science.sciencemag.org/content/296/5570/1029
[3] Ibid.
[4] T. Cook, “CNBC Interview With Tim Cook.” CNBC (2019), www.cnbc.com/video/2019/01/02/apple-tim-cook-revenue-stocks.html
[5] H. Grossman, “Google vs. Death”, Time (2013), content.time.com/time/subscriber/article/0,33009,2152422–1,00.html
[6] N. Carr, “Is Google Making Us Stupid?”, The Atlantic (2018), www.theatlantic.com/magazine/archive/2008/07/is-google-making-us-stupid/306868
[7] Ibid
[8] J. Cascio, “Get Smarter”, *The Atlantic(*2014), www.theatlantic.com/magazine/archive/2009/07/get-smarter/307548/
[9] C. Metz, “Making New Drugs With a Dose of Artificial Intelligence”, *The New York Times(*2019), www.nytimes.com/2019/02/05/technology/artificial-intelligence-drug-research-deepmind.html
[10] A. Kundaje, “Anshul Kundaje: Machine Learning to Decode the Genome”, The Future of Artificial Intelligence (2019), www.youtube.com/watch?v=lX76DzZdjvQ
[11] “What Are Genome Editing and CRISPR-Cas9? — Genetics Home Reference — NIH”, U.S. National Library of Medicine, https://ghr.nlm.nih.gov/primer/genomicresearch/genomeediting
[12] I. Sample, “Google’s DeepMind Predicts 3D Shapes of Proteins” The Guardian (2018), www.theguardian.com/science/2018/dec/02/google-deepminds-ai-program-alphafold-predicts-3d-shapes-of-proteins
[13] M. Coren, “Foldit Gamers Solve Riddle of HIV Enzyme within 3 Weeks”, *Scientific American (*2011), www.scientificamerican.com/article/foldit-gamers-solve-riddle
[14] I. Sample, “Google’s DeepMind Predicts 3D Shapes of Proteins” The Guardian (2018), www.theguardian.com/science/2018/dec/02/google-deepminds-ai-program-alphafold-predicts-3d-shapes-of-proteins
[15] A. Ng, “IBM’s Watson Gives Proper Diagnosis for Japanese Leukemia Patient after Doctors Were Stumped for Months”, New York Daily News (2018), www.nydailynews.com/news/world/ibm-watson-proper-diagnosis-doctors-stumped-article-1.2741857