All of the AI in Biopharma events are designed to be interactive so that attendees can easily get their questions answered. In addition, attendees are able to engage with each other during the events and on the AI in Biopharma digital community platform year-round. On September 15, 2020 the second of a three-part series of events was held. Below you will find selected talking points that I found to be thought provoking and valuable insights. Please note that these are summaries and not verbatim comments.
Topics discussed at the Data Management Strategies to Effectively Use AI Tools include:
Part 1: Convergence and Critical Mass – The Fusion Moment for Biopharma and Data
- A Moral Imperative to Share Data and Impact Patients Faster
- Proper Attribution of Contributions
- Will People Use the System?
- Grassroots and Top-Down
- Reproducibility and Standardization of Data
- Unlocking the Information That Makes Cures Possible
Part 2: Data Curation and Privacy
- Roles and Responsibilities
- Governance Areas Where AI Can Be Leveraged
- Better Data Curation Leads to a Better Patient Journey
- Data Sharing Platforms and Privacy
- The Patient as Partner is Missing in Downstream Uses of Data
- Defining Personal Data
- Data Stewardship
Part 3: Using AI to Generate Patient Data – How Digital Twins are Shaping Clinical Trials
- Background on Digital Twins
- The Unlearn Approach to Using AI
- Data Quality vs. Data Quantity
- The Promised Land of Clinical Trials
- Digital Data Collection
- Building a Digital Twin
- The FDA Response to Digital Twins
- Continual Validation and More Efficient Trials
Part 4: Drug Development Using RWE, Integrated Research Platforms and Advanced Analytics/AI
- An Integrated Research Platform and Portfolio of Therapies
- The Status Quo is a Harsh Mistress
- No Excuses
- True Innovation Will Require an Adaptive Approach
- Data Standards and A Common Language
- The APANDEMIC Initiative Goals
- Pending Legislation
- Closing Thoughts
- Thank You to Event Sponsor Unlearn.AI
- October 26th Digital Summit: AI Case Studies in Discovery & Clinical Development
- Thomas Bock, MD, MBA, Founder & Past CEO, HeritX, Inc., Chair, Healthcare Advisory Board, Columbia Business School; Former SVP, Global Head of Medical Affairs, Alexion; Co-Founder, The APANDEMIC Initiative
- Charles Fisher, PhD, Founder and CEO, Unlearn.AI
- Belen Fraile, MD, MSc, VP & Head, US Oncology Strategic Data & Digital, Novartis Oncology
- Harry Glorikian, General Partner, New Ventures Funds
- Anna Langhorne, PhD, Esq, Global Data Privacy Counsel, Regeneron Pharmaceuticals, Inc.
- Brian S. Martin, Head of AI R&D Information Research, Research Fellow, AbbVie
- Ülo Palm, MD, PhD, Chief Medical Officer and Co-Founder, Ordaos Bio; Co-Founder The APANDEMIC Initiative
- Peter Pitts, President and Co-Founder, Center for Medicine in the Public Interest; Former Associate Commissioner for External Relations, FDA, Visiting Professor, Université Paris Descartes Medical School, Chair of Innovation Management & Healthcare Performance
Part 1 Interview
Convergence and Critical Mass: The Fusion Moment for Biopharma and Data
Brian S. Martin, Head of AI R&D Information Research, Research Fellow, AbbVie
Brian was the opening speaker for the day and spoke passionately about the need for pharma companies to change their organizational culture so that ownership of information is no longer an internal hurdle to overcome on the path to improving productivity and impacting patients. This theme was echoed in the other presentations that followed Brian’s.
A Moral Imperative to Share Data and Impact Patients Faster
A core project at AbbVie is the building of a substantial knowledge graph that they hope will become knowledge and data that can be shared with a broader group of people in the life science ecosystem. Ultimately the goal will be that a large group of people will be able to mine that data for insights to help everyone get to cures, and to impacting patients, faster.
Proper Attribution of Contributions
How attribution of work on data is done is an important component of the system AbbVie is building. A key function of the technology platform is the ability to maintain the lineage, the tracing of the knowledge, back to its source data and the data back to its source operational systems. If a certain piece of knowledge or insight about a certain compound is discovered, the knowledge of that can be attributed back to the data for attribution for everyone who has contributed along the path.
Will People Use the System?
The carrot to get people to use this system is having all of the data in one place and having the “janitorial work” to keep the data clean and normalized is part of the platform so that saves people from having to do that work. Users are also getting much more compute power than they might be able to get if they are not willing to work within the platform capability.
Grassroots and Top-Down
As part of the convergence initiative, there is an organizational structure being developed that is grassroots and top-down to drive the success of the initiative.
Reproducibility and Standardization of Data
Ultimately AbbVie hopes to publish out the ontology built into the system to enable the reproducibility and standardization of the data. The goal is that this will be available within AbbVie, but also will be made available to other groups so they can learn from the work that AbbVie has done and understand how the data and knowledge that they are seeing is connected and structured. This is a long process that is just getting underway with many legal and other hurdles to be overcome.
Unlocking the Information That Makes Cures Possible
The biggest question around this project is how to make this valuable for the scientists who are working on unlocking the information that makes cures possible. This needs to be done at greater scale, with more precision, and in new ways to be a success.
Part 2 Discussion
Data Curation and Privacy
Belen Fraile, MD, MSc, VP & Head, US Oncology Strategic Data & Digital, Novartis Oncology
Anna Langhorne, PhD, Esq, Global Data Privacy Counsel, Regeneron Pharmaceuticals, Inc.
Roles and Responsibilities
Anna’s role is to make sure that Regeneron handles personal data in compliance with global privacy laws and regulations. Personal data comes from many places including patients in clinical trials, employees, and marketing programs. Anna sees her role as empowering scientists and infusing the process with privacy-by-design so that they can get to a privacy-compliant place. This mentality is good for scientists and for patients.
Belen’s team focuses on processes and technology but also ensures that the dialog with patients is done correctly and is built on a model of patient-centricity.
Governance Areas Where AI Can Be Leveraged
The nature of data and the consents under which it was collected all impact the governance of that data. Leveraging data and the context around it, who will be touching the data, and where has it come from, are governance areas where AI can be leveraged effectively. Improving traceability, visibility and transparency to patients are important areas where AI can also be useful.
Better Data Curation Leads to a Better Patient Journey
Data is a mirror of the patient journey and in many cases the system and journey are broken and are more complex than necessary. What is the role that pharma can and should play as co-creators of patient-centricity in the ecosystem? Better data curation is where a better patient journey needs to start.
Data Sharing Platforms and Privacy
There needs to be a culture shift in thinking about the treatment of data. From a privacy perspective, Anna is very supportive of data sharing platforms, but the sharing of data needs to happen in a privacy compliant way.
The Patient as Partner is Missing in Downstream Uses of Data
Patients want to share information, but they don’t have visibility into downstream uses of their data. Poor communication with patients and not explaining what happens with data in a meaningful way is a concern. The patient as partner is missing.
Defining Personal Data
How is personal data defined and how do you manage anonymization when different countries, regions and professionals have different definitions of what this looks like? Thinking about data retention is another key element of empowering patients in a way that they can see and understand how and when their data is being used. When data is shared on platforms that use good data governance practices, this can create a better equal opportunity environment not just for scientists but also ultimately for patients.
AI will provide outputs based on models you have running so worrying about different regulatory perspectives can be solved within the system itself. By setting up the right data stewardship, you are doing the work that needs to be done to adapt for different uses and regulatory environments. The right tagging, data management and governance are the critical foundation. Getting information to patients is about processes and governance and not systems or technology.
Part 3 Interview
Using AI to Generate Patient Data: How Digital Twins are Shaping Clinical Trials
Charles Fisher, PhD, Founder and CEO, Unlearn.AI
Background on Digital Twins
Unlearn.AI has developed a computer simulation of a patient that enables you to ask questions about what would happen to a patient if they received a particular treatment. This is compared to a control treatment, which is usually a placebo to maintain the elements of a RCT. The digital twin enables more efficient trials because fewer people need to be taking the placebo. Unlearn uses data from previously completed trials and EHR data and can create ML-based simulators to learn how to create a simulation model of a particular patient. Work is being done now on data from Alzheimer’s patients.
The Unlearn Approach to Using AI
Unlearn uses AI by taking in a diagnosis and then the output is an image. The result is being able to say, for example, this is representative of an X-ray that is consistent with a particular diagnosis. This is a different approach to using AI than is being used in many pharma applications. Validation in many ways is more difficult than a standard ML use case.
Data Quality vs. Data Quantity
As a general rule, data quality is vastly more important than data quantity. Unlearn is working to create detailed simulations of patients so trial data is more useful than EHR data. EHR data can give you a diversity of data, but a much lower quality of data. Trial data tends to be more homogeneous but higher quality.
The Promised Land of Clinical Trials
The promised land of clinical trials would be to enroll instantaneously with half as many people and take half the time and cost half the amount they currently cost. They should provide information about diverse patient populations as well as individual patients. The way to get there is to start by integrating different data sources including EHR data and trial data.
Digital Data Collection
Any data that you can measure can be included in the patient’s digital twin. The short-term situation with clinical trials has been to shift to digital data collection because of the pandemic. If that turns out to be very successful, that could be a permanent change.
Building a Digital Twin
To build a digital twin there is a process of training the simulation model on historical data before a trial event starts. That process bottleneck is in integrating and cleaning the data to get it ready to use. Unlearn is creating high quality data sets so that is the bulk of the effort, sometimes taking a few months. The actual process of creating a digital twin once you have the high-quality data model is not that complicated. Unlearn is essentially generating medical records that can then be integrated into current workflows in clinical trials in a very standard way. The process of training the model is long and difficult but the process of using it is a lot easier.
The FDA Response to Digital Twins
Interactions with the FDA has been very supportive of digital twins so far. Ultimately the regulatory response will depend on the use case. One approach is that digital twins can be used to reduce the size of the placebo arm. This can now be a more efficient trial that includes randomization and you get additional power by leveraging AI and ensuring rigorous results.
Another approach could be eliminating the control arm and running a computer simulation of the control group. This is not something that would be done in later stage trials, but could be effective for early stage proof-of-concept studies.
Continual Validation and More Efficient Trials
One Unlearn model is based on one data set that eliminates re-validating the model with every trial. This is a continual validation model that can then be used across clinical trials for a particular indication.
Part 4 Panel
Drug Development Using RWE, Integrated Research Platforms and Advanced Analytics/AI
Thomas Bock, MD, MBA, Founder & Past CEO, HeritX, Inc., Chair, Healthcare Advisory Board, Columbia Business School; Former SVP, Global Head of Medical Affairs, Alexion; Co-Founder, The APANDEMIC Initiative
Ülo Palm, MD, PhD, Chief Medical Officer and Co-Founder, Ordaos Bio; Co-Founder,The APANDEMIC Initiative
Peter Pitts, President and Co-Founder, Center for Medicine in the Public Interest; Former Associate Commissioner for External Relations, FDA, Visiting Professor, Université Paris Descartes Medical School, Chair of Innovation Management & Healthcare Performance
An Integrated Research Platform and Portfolio of Therapies
We need to learn from patients in an iterative way. The current system is not working efficiently. We need to bring all data sets together onto one platform with an adaptive approach to make trials more informative. An integrated research platform needs to be the first step in effectively using AI. COVID has created more collaboration than existed before. Using AI and advanced analytics can also help to move forward faster with portfolios of therapies instead of focusing on one therapy at a time.
Utilization of an integrated research platform will require much more collaboration between research institutions, pharma companies and others. If Facebook can have millions of people on a platform, this should not be impossible when it comes to improving health and outcomes.
The Status Quo is a Harsh Mistress
The status quo is a harsh mistress. We need to look at innovative ways to move forward and there are multiple ways to get to the goal of speeding up trials. The concept of using a parallel track of RCT and RWE with AI and other tools is the smartest way forward. The only way innovation and improvement will happen is if people look at new ways to integrate, use and validate RWE in new products. The old paradigm and historical gold standard of trials need to be adapted to be able to make progress. We need to put a stick in the eye of tradition.
We now understand that we need to recruit a diverse population for trials. Genomic data has highlighted this need. This understanding is progress, but we need to do more. The FDA needs to better understand that advanced technologies should be put to use beyond oncology.
The FDA should not be used as an excuse or barrier for making change. Academia, FDA, and pharma all need to change their thinking and forget about the status quo and old gold standard. We need to think about what we need to do now to be successful with the current tools that we have access to.
We are now at a crossroads. One road leads backwards to the known and the comfort zone and the other road leads forward to innovation.
The current pharma model is most likely not one that we will still have twenty years from now. The start-ups of today could be the leaders of tomorrow. These companies have visions for reinventing drug discovery and drug development, and they shouldn’t be dismissed by big pharma. These companies will make a difference.
True Innovation Will Require an Adaptive Approach
AI in the field of drug discovery is very promising. We are now starting to see more uses for AI in trials and COVID has propelled the idea of using a more adaptive approach to trials. We now need to move to using RWD and focus trials on the population you are trying to serve and use AI tools to sort through the disparate data to discover patterns that only a computer to discover. True innovation will require adding an adaptive approach to trials that is not currently in place.
Data Standards and A Common Language
We need to have a standard for all data that is a common language. We can’t have every company speaking a slightly different language. Agreement for a common data standard need to be set.
The APANDEMIC Initiative Goals
At the beginning of the pandemic a network of healthcare professionals and companies came together to form the APANDEMIC Initiative. One of the primary goals of the initiative has been to create an integrated research platform. The second goal is to work with regulatory agencies in the US and in Europe around the topic of data validation. The third goal is to move toward eventual approval of treatments coming from a variety of data sources. This would require external reviewers and regulators to closely monitor patients over time.
The Promising Pathways Act is a legislative initiative around these issues that has been introduced in the Senate for people interested in legislative issues around these topics.
If we don’t take the COVID-19 situation and use it in positive ways to advance the argument, then shame on us.
New methods like AI are going to change things. The new generation, the twenty-somethings, will move beyond the old paradigms.
COVID has shown us that the traditional model where you set up a hypothesis and wait four years until you see results does not work for every situation so we need to seize the opportunity and get the innovation done.
Thank you to AI in Biopharma Sponsor
Unlearn has developed the first machine learning (ML) platform for creating Intelligent Control Arms with Digital Twins through its proprietary DiGenesis™ process, allowing drug developers to dramatically reduce therapy development time, while lowering the risk of trial failure, thereby increasing confidence in clinical trial results. Unlearn is working closely with biopharmaceutical and medical device companies as well as regulators to ensure its methods meet the highest scientific and regulatory standards. Visit https://www.unlearn.ai or follow @UnlearnAI on Twitter, @unlearn-ai on LinkedIn.
AI in Biopharma Digital Series
Connecting biopharma’s science and technology is what we do at AI in Biopharma. To continue this important discussion, with a focus on AI Case Studies in Discovery & Clinical Development, register for the upcoming interactive digital summit on October 26th. Speakers include thought leaders from: Envisagenics, Systems Oncology, Novartis Institutes for Bio Medical Research (NIBR), Genentech, Genialis, H1, AbbVie, Bristol-Myers Squibb, GSK, TransCelerate BioPharma, Inc., Genuity Science, and moderator Harry Glorikian, General Partner, New Ventures Funds.
If you have any questions or suggestions about the digital event series, if you are interested in speaking at an upcoming webinar, or would like sponsorship information, I would love to hear from you. You can contact me directly at: firstname.lastname@example.org.