The Center for Data Innovation spoke with Sohini Chowdhury, deputy chief executive officer of The Michael J. Fox Foundation for Parkinson’s Research (MJFF), based in New York. Chowdhury discussed how MJFF makes valuable data available to researchers, as well as the role data collected from Internet-of-Things devices can play in leading to new treatments.
This interview has been lightly edited.
Joshua New: Why does Parkinson’s disease research stand to benefit so greatly from the increased use of data and AI?
Sohini Chowdhury: That’s a great question and what I would say is that my answer is twofold. One is that the answer is generalizable beyond Parkinson’s disease (PD). And then there is a component that is very PD-specific.
When we think about Parkinson’s disease, like many diseases, we say “if you’ve met one Parkinson’s patient you’ve met one Parkinson’s patient.” There’s an enormous amount of heterogeneity in this disease. That’s not specific to PD, which is why some of this is generalizable beyond Parkinson’s. When you’re able to look at large amounts of data from thousands of patients you begin to understand what is noise and what is a real signal that you can then build treatment or drug therapies around.
I think the beauty of novel ways of collecting data that exist now and emerging techniques that allow us to analyze data in ways we haven’t before is that we can really begin to tackle the problem. We can tackle the challenge of disease heterogeneity in a way that we haven’t before.
What’s particularly specific to PD and why it’s so exciting when one thinks about novel data collection techniques and AI is that, as a disease, Parkinson’s lends itself well to capturing different components through novel data capture techniques. By that I mean the motor aspects of the disease, the rigidity, the stiffness in walking, or the balance issues. These are things that actually can be measured through sensors whether in a clinic or with sensors that are readily available in retail devices like an Apple watch or FitBit. What makes Parkinson’s very unique is that you can capture so much of this information related to movement, which is such a core aspect of PD, on an ongoing basis so that you’re really capturing a full spectrum of the disease experience that an individual is undergoing on a daily, weekly, or monthly basis. It allows us to hone in on aspects of the disease in a way that we haven’t before because it’s actually now quantifiable and trackable.
The implications of novel data collection and AI are really exciting for many diseases, but particularly so for Parkinson’s.
New: What is the Fox Trial Finder? What kind of challenges does it solve for the clinical trial process?
Chowdhury: Fox Trial Finder is what I like to call the “Match.com” of the PD clinical trial ecosystem. It pairs interested volunteers with clinical research opportunities by using a backend algorithm to look at the characteristics of an individual and match them with the inclusion and exclusion criteria for different research opportunities. So you’re not just getting a registry or a roster of research opportunities, but a more tailored list that provides you with opportunities that match you and your interests. What this tool is trying to solve is basically the problem of recruitment, which is a problem with research into many diseases. There are tons of research opportunities out there but we’re struggling to find participants for those opportunities on a timely basis. Fox Trial Finder provides an easy platform to help make these matches faster.
New: MJFF launched its Fox Insight research portal in April. What is the purpose of this portal?
Chowdhury: The Fox DEN, or Fox Data Exploration Network, is the Fox Insight portal through which researchers can access a dataset from our Fox Insight study. Fox Insight is an online clinical study that invites participants, whether they have PD or not, to provide information about their lived experience. So if you don’t have PD, you can list any medications you might be on, any vitamin supplements you take, how much exercise you do, and things like that. If you have Parkinson’s, you provide information about your medications, when you were diagnosed, and your symptoms.
The point of Fox Insight is to help support the development of what we like to call patient reported outcomes, or PROs, which tell us what matters to patients. Increasingly, the concept of PROs is becoming more important to regulators when they’re looking to evaluate a drug and trying to understand the effect it has on a disease. They want to see not just a biological effect that the drug may have, but also whether the drug is having a functional impact on an individual’s life. Is it actually helping that individual in his or her day-to-day existence with this disease?
Fox Insight collects this information directly from patients and participants without the filter of a clinician. The study was launched almost two years ago now and we have to date more than 38,000 participants, 75 percent of whom are PD patients. The Fox DEN research portal allows researchers to mine this data to look for new research questions.
New: What is the Parkinson’s Progression Markers Initiative (PPMI)? Why are “progression markers” so important?
Chowdhury: The PPMI studies several different types of cohorts. It includes what we call idiopathic Parkinson’s disease patients, which are patients that developed Parkinson’s without any known genetic mutations. It includes PD patients with genetic mutations. It includes control participants, who are individuals without Parkinson’s. And it also includes individuals who don’t have Parkinson’s but do have genetic or other risk factors that increase the possibility that they may develop PD. The study has followed these cohorts on a longitudinal basis with an enormous amount of data being collected. This data includes clinical data, different imaging modalities, biological samples, cerebrospinal fluid, and more.
In the nine years since this study was launched, the PPMI dataset has become the deepest phenotype dataset that exists for Parkinson’s disease. It has about 1,400 individuals in the study and the amount of data on each individual is huge. It allows researchers to understand the progression of the disease beginning with risk factors or diagnosis, all the way up to having the disease for four or five or more years. This data can help us develop the biomarkers that can determine whether a drug is slowing or stopping the disease’s progression. It also allows us to get smarter about identifying who may be the right participants for the right clinical trials, as certain drugs may be targeting certain genetic mutations.
This is an unbelievably unique open access resource. Researchers can download this data and can request access to biological samples. To date, we’ve had almost four and a half million downloads of the dataset and we’ve had about one hundred and fifty requests for the samples. So it has really become a research engine for the Parkinson’s community.
New: Has the MJFF’s efforts to make data more available for research lead to any new treatments for Parkinson’s, or is it too early to tell?
Chowdhury: I would say that it’s still early. Sensors and other novel data collection have only come into widespread play, I would say, over the past two or three years and, as you’re probably well aware, drug development is a decades-long process. It’s a little bit too early to create a causal link between our data and “X” new treatment for PD.
That being said, one of the really interesting things about PPMI is that we have recently initiated new kinds of digital data collection for these cohorts. We’re collecting data through collaborations with Roche and Verily in two ways. With Roche, we are using a mobile phone app that they’ve developed for their phase 2 Parkinson’s clinical trials that will be collect a combination of passive data, meaning data from while the phone is being held or in a person’s pocket, as well as structured data from daily surveys and tests.
For the collaboration with Verily, they have provided us with their smart watch to passively collect data as well. For the first time, this passive collection will allow us to correlate data from different types of individuals with PD and individuals at risk of developing Parkinson’s over the course of the study. We’ll be able to actually pair this data with imagery, biological data, clinical data, and more.
I think that is super exciting and will lead to huge avenues of opportunity and acceleration in drug development.