Manifold, a data platform startup company that supports clinical researchers, recently announced that the American Cancer Society (ACS) will leverage its platform to modernize its population research study management. Manifold CEO Vinay Seth Mohta recently spoke to Healthcare Innovation about the pain points in clinical research his company is addressing.
Manifold announced its $15 million Series A funding in April this year, and also announced partnerships with Indiana University Melvin and Bren Simon Comprehensive Cancer Center and Winship Cancer Institute of Emory University. Manifold said it is supporting the American Cancer Society’s VOICES of Black Women study by streamlining all aspects of population research, including simplifying the participant experience for surveys and collection, and enhancing how study managers design dynamic surveys, engage with participants through personalized e-mail and SMS outreach, and monitor participant data over many years.
Healthcare Innovation: Before we talk about what Manifold does, could you tell us a little about your background before co-founding Manifold. I saw that you were involved with Kyruus and also were an adviser to some other startups like Firefly, is that right?
Mohta: I came up through technology in computer science and electrical engineering and then really been working in software that deals with data for most of my career. I started early at MathWorks, which is software that's used for a scientific computation and engineering, and then really got bitten by the startup bug in the late 1990s and pretty much never left that world. I have progressively gone from being very technical to thinking of myself as sitting at that intersection of business and technology, and how we can best bring this deep technology we have to work with data to different areas.
HCI: What was the origin story of Manifold? I looked on your web site, and it sounded like the initial focus was to become an AI lab, and then you changed directions?
Mohta: Yes, that's exactly right. Several of the co-founders have a connection to MIT. We had the idea of an AI lab that is pragmatic and takes the research and really brings it to application. That's where we were seeing the initial opportunity. And then because of our connection to healthcare, as we were looking to apply AI, we began to see that there were a lot of workflows that were broken — particularly, as we started to narrow in on research. That's what led to some of the insights for what is now a software platform.
HCI: A lot of your solutions are in cancer data management and cancer data sharing. What are some of the limitations that cancer researchers experience that these solutions can help them overcome?
Mohta: Broadly in clinical research, which also includes epidemiology, there are just a lot of manual processes. In the same lab, you'll have a researcher cutting genes with CRISPR, and then turn around and use spreadsheets and Outlook to manipulate the data.
There's a lot more data about humans that's available today. Twenty years ago, we had paper medical records, but today, you can access your health record electronically. You have digital pathology inside these organizations, and your radiology is electronic. Then, of course, genomics is becoming standard of care in cancer. These are things that you can't deal with manually; you can't move it around in spreadsheets. What we started to see was so much manual work happening that was really taking away time from researchers actually doing clinical research and science. Disparate data silos and manual work are the themes we saw.
After talking with customers at cancer centers and large scientific organizations like the American Cancer Society, we decided that cancer seems to push both the science and the technology the hardest first, so that's where we thought the pain was. And what happens in cancer today is what's likely to happen in other disease areas in five to 10 years. So we felt like it's a great place to start partnerships.
HCI: Is part of this effort about gathering multimodal data — genomics, imaging and EHR data — from where it might sit in an academic medical center and bringing that all together onto one platform? Is that what we're talking about as far as breaking down silos? And maybe across institutions as well?
Mohta: Yes. As an example, in our work with the American Cancer Society, they have experience working with these 100,000-plus people cohorts over multiple decades. And one of the big pain points they surfaced was the downstream data management, data analysis, data science, and external collaboration — all of the things you just called out. But also, are participants actually engage with the study and how do you keep them? How do you enroll them? How do you manage their concerns? The study management team has a group of people that will take a phone call from the participant who says, ‘Hey, I don't understand this question in the survey. What should I do here?’ All of those workflows are also very manual.
A researcher might see that the last 100 patients responded differently to some of the treatments, and want to reach out to them and ask them a few questions. The tools for those kinds of work are also quite primitive. So if you think about that end-to-end journey from engaging participants, gathering data from them, and gathering data from other digital systems, which are the silos, and then putting it all together for both internal analysis and external analysis, that’s exactly the pain points we saw and how we've approached the product.
HCI: How does AI come into the platform? What does it enable?
Mohta: Our goal, big picture, is to make it 10 times cheaper and 10 times faster to stand up clinical studies, run clinical studies, and run epidemiology studies. I'll be the first one to say we're not there today. But we're well on our path to doing that. Because, again, if you add up the costs of all of this manual work people are doing, there's so much opportunity to go faster. So that’s the motivating theme for us. We have a platform so that people don't have to build all of this internally with their IT organizations. We've productized it as opposed to everybody having to do it bespoke. So that's one place you get leverage.
The second place is where we get to AI. And there are really two places I think about AI. One is the data management piece. How do you make this data link up to the same participant? How do you take genomic data and make it be more accessible to a clinician scientist or an epidemiologist who doesn't know how to write code? All of that data cleansing, data manipulation and harmonization is one of the big places we're applying AI to link these datasets together in a useful way.
I think the other one is on the user experience side. And this is definitely one you see other people doing in other domains. But once you have all of this data together, can we make it easier to access the data as well? For a clinician who is seeing patients, if they have some questions about what were the outcomes for my population, there is a very high barrier. They had to go request programmer time; they had to figure out what systems data was in and get approvals. Can we take that and bring it to more traditional interfaces that are graphical, like dashboards and search interfaces, and then take it to a conversational interface?
To us, it's really reducing the cost through using AI to make some of the data pieces easier to do, and then increasing the value of that data now that you've done that work, where a lot more people can do research on it, ask questions on it, etc.
HCI: When we think about all of the sophisticated cancer research centers across the country, have a lot of them built this capacity in house or are there competitors to Manifold that have already helped them do that?
Mohta: I think we are first of a kind in terms of this going from the participant experience all the way to external collaboration. These institutions might have a point solution for how they survey participants today, for example, or they might have a point solution for how their genomics data is available to their bioinformaticians. Or they might have a point solution for their digital pathology system. They might provide some tools to work with digital pathology images, but we are taking that end-to-end view, and then putting the multimodal picture together.
I was talking to the chief research informatics officer of a prominent institution last week. I said, ‘You’re a well-known, leading institution. Surely you've solved some of these problems.’ And they said, ‘Well, we've solved it in pockets at the level of a lab. But there's really nothing at a production scale, enterprise-grade, fully supported.’ I would say that's pretty consistent with the theme that we've seen is that there are a couple of pockets where researchers have that informatics capability. They were at the leading edge, and they had to do it for themselves, so they did it as part of a grant. But really, again, this idea that you can make it available to a broad part of the the ecosystem at a much lower cost, and much more quickly is where we see the opportunity, rather than people thinking about trying to build it themselves, which is expensive and takes a lot of time.
HCI: Manifold also has worked with the cancer center at Indiana University. Can you talk a little bit about what they're doing?
Mohta: In their case, they had a similar theme, where it was the biobank and the biospecimen core, which provides some shared services to the rest of the cancer center. The change for them, thematically is very similar, where 10 years ago, somebody would ask, do you have pancreatic cancer tissue? Today, when a researcher comes to them, they ask, do you have pancreatic cancer tissue that has a BRAF G608H mutation where patients already received immunotherapy? They're also seeing this multi-modality come together. Where tissue’s been sequenced, there's RNA data, there's DNA data, there's imaging from pathology, and researchers come to them looking for the answer to that question of: Do you have data? Do you have tissue? They wanted to provide those tools around that integrated view of the data, and then also are starting to say to the researchers, ‘Hey, you can come and explore this data on your own; you can check out what we have, and get to the heart of the scientific question. And that's what people are spending time on, rather than pulling data from one system, then going to a second system, going to the third system, with spreadsheets or custom code, and then trying to answer researchers’ question. So you really take those early researcher questions from weeks to seconds or minutes, and really transform the productivity of the organization overall.
