The real impact of the All of Us Program

Today, May 6, happens to be the national launch day for NIH's All of Us program. This is a program that focuses on precision medicine - an emerging approach for disease treatment and prevention that takes into account individual variability in genes, environment, and lifestyle for each person (definition from NIH).

Precision medicine is not a new concept. If you google for it, you can find several documented cases, especially in cancer treatment where genomic information of a single patient was used to specifically target his or her disease successfully. The Director of the All of Us program, Eric Dishman happens to be such a case and is the kind of leader the program needs to make sure everyone can make use of it.

Using the genomic data to precisely understand and define medical treatment for an individual seems logical. But how can we generalize this and make its reach wider? How can the treatment of one person at one end of the world have a positive impact on a patient at another end? How can we connect the dots? Can we broker the innovation produced by Facebook to socially engineer the ad targeting into our own healthcare? Can we target medicines or health treatments to an audience instead of ads?

The answer is of course a resounding YES. It is resounding enough to reveal that All of Us is not the first program to attempt to do so. So why would it be more successful than other data collection programs? It has many good things going on for it. Besides large national funding and a passionate team that is behind it, it attempts to engage community, capture diversity and use all available technology.

As I wrote in my previous article, the power of using data to help complement today's medicine is huge. The All of Us program extends the scope of this power to come from a much wider data source. This is the data collected from a lot of people and a lot of types of data besides just genomic data. It targets not just patients but also healthy people. So why is all of this necessary if we can just take someone's genome and use it to tweak their treatment. Think of what Facebook does by collecting large amounts of data from a large set of people having diverse behavior. Think of how Alexa learns to do some smart things by learning from questions asked by a large number of people. Perhaps a quick lesson from machine learning would be useful here.

Machine learning from data involves looking for patterns when trying to see why certain choices or events or input parameters led to a set of outcomes. Finding the patterns needs diversity of data, else machine learning can go very very wrong. Data which is skewed towards certain outcomes can lead to very poor models. In addition, the presence of redundant inputs or the absence of key inputs can lead to poor models too. Having too little data is also not helpful and can lead to failed models in real world. Data scientists often combine the results of many models that are based on many sets of data to build a more real world robust model. 

That is precisely why the program attempts to collect a diverse set of data from a diverse set of healthy and non-healthy individuals and make it available to a diverse set of researchers. This could fundamentally change how medicine works in a few years. Think of how much machine learning technologies have impacted our day to day work in the last ten years. And how cheap it has become to research and deliver the benefits! Very simple health related outcomes could be achieved in the next couple of years itself. More complex cloud based solutions that use this data to buy health benefit programs "as a service" could evolve in few years!

So really what is the All of Us Research Program?

Saturday today and I was stopped in my tracks to finish my chores several times. Tomorrow (Sunday) is the big launch. The big launch of what, they asked? They were messaging me all over the place on social media. Just google it, I said.

It finally is here - the big day, May 6 where NIH is engaged in the national launch of the program that has been in gestation so far for several months with thousands of beta participants. The program aims to enroll one million participants to volunteer their data and other information so researchers can use it to improve medicine.

I had no idea of this program until the middle of last year when I came across Vibrent (based in NoVA) and met its CEO, Dr Praduman Jain who later went on to explain the benefits and goals of Precision Medicine very passionately to me. And then it finally dawned on me. 

Well that was then. Fast forward almost four months and here we are. We have been so busy building up the functionality and infrastructure for the program that it has been all to easy to forget what the program is about!

A few months ago, I read an article by Chris Anderson, former editor in chief of Wired magazine, that I thought applied very well here: “The end of theory: the data deluge makes the scientific method obsolete” (http://archive.wired.com/science/discoveries/magazine/16-07/pb_theory/). As the title indicates, Anderson asserted that in the era of petabyte information and supercomputing, the traditional, hypothesis‐driven scientific method would become obsolete. No more theories or hypotheses, no more discussions whether the experimental results refute or support the original hypotheses. In this new era, what counts are sophisticated algorithms and statistical tools to sift through a massive amount of data to find information that could be turned into knowledge (source: NIH).

The All of Us Research Program can be instrumental in this process. Today, scientific research happens in the laboratory. A sample or specimen is observed, tests developed and then trials are done before getting approvals and mass rollout. This is still the wide majority of cases. Behavioral scientists have obviously been using data for their research regularly and there are few cases where actual patient data was used to build out a specific medicine formula that was then given to them. The All of Us Research Program has the potential not only to lower the costs of research in medicine but also to increase the efficacy of medicine.

Consider the long tail problem in developing medicines. The All of Us Research program can have a huge impact on solving this issue. If I was to give an analogy, this program has the ability to do what fiber optic cables did for the internet. I am not talking about the stock market crash of 1999 ! I am referring to the immense large possibilities of breaking down and then solving a very complex and large problem. Imagine what kind of collaboration you could get if this kind of data is available across national boundaries and what if it can be used and compared on demand by researchers all over the globe? This could completely democratize the problem of finding cures for diseases and help develop a very competitive market for medicines in the next 20-30 years. Hopefully that happens in my lifetime!

PS: This blog describes my person viewpoint on the All of Us Program. For the official definition or for signing up for the program, please visit the NIH Join All of Us website or download the apps from the App Store or Play Store. If you are interested in the crazy engineering that goes behind building this, join Vibrent Health. We are on the lookout for passionate (and crazy) engineers that want to change the world.