Originally published Jul 16, 2020

by Elizabeth Cutler, Co-Founder, AI in Biopharma

As a Co-Founder of the AI in Biopharma Digital Event Series, my goal is to provide the most compelling, engaging, and helpful information to biopharma professionals who are using AI/ML tools. 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 July 14, 2020 the first 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 Making AI/ML Tools Scalable Across the Drug Development Lifecycle to Impact Patients include:

  • Data Creators and Data Consumers
  • Do Things Better and Do Better Things
  • An Operational Quick Win Using RPA
  • Data Sharing is Essential
  • Noisy Data
  • The Impact of AI/ML Tools in the Value Chain
  • Partnerships Always Go Back to People
  • Final Thought and September 15th Digital Summit

Participating speakers

  • Christian Baber PhD, Global Head of Scientific Informatics, Takeda
  • Mark Borowsky, Ph.D., Global Head of Information Products and Data Science, Novartis Institutes for BioMedical Research (NIBR)
  • Harry Glorikian, General Partner, New Ventures Funds
  • Niven R. Narain, CEO, BERG
  • Andrew A. Radin, CEO, twoXAR Pharmaceuticals


Part 1

Interview with Takeda’s Christian Baber: Unlocking the Power of Data

Decision Support
Christian’s focus in on decision support which he defines as using data in any way possible to support the scientists making decisions. The biggest challenge is around the diversity and flexibility of data in early stages of drug development. There is also a lack of negative data, which is important to include as you use data to build effective models. The best way to think about AI/ML technology is that it provides augmented intelligence to help scientists make decisions and do their job better and faster.

Agustin Marquez
The data creators are often not the same people as the data consumers. What Takeda is doing is recognizing that fact and bringing the scientists into the process early to contribute to decisions around labeling and organizing data.

Staffing Models
Takeda uses a federated model of staffing which means that IT is centralized, Christian has a decision-support team, and within R&D there is a central data science institute. Each group views the other as a partner that provides complementary work and they avoid duplicating efforts. Hiring needs to be balanced between people with domain knowledge and people from other domains. A flexible, innovative, and adaptable mindset in people you hire is key for roles working with AI/ML tools and predictive models.

Do Things Better and Do Better Things
The data used to build technologies needs to have one of two goals, either do things better or do better things. Doing things better is the easy part, it’s about operational efficiency and quality. Make it easier for people to apply meta-data and make sure people only have to enter data once.

An Operational Quick Win Using RPA
RPA has mainly been used at Takeda on the reporting side at this point. They are now starting to look at it in the non-clinical and clinical space. If you go beyond RPA to other forms of automation, there are also efficiencies in quality of data and fewer people needed for manual data checking for errors, as an example.

Data Sharing is Essential
In general, data becomes less valuable and relevant over time. The power of ML and analytical methods grows exponentially with the number of data points, so data sharing is essential. Data sharing between companies is very difficult but there have been good efforts made during the pandemic. Good work has also been done in the past in the tox area for the greater benefit of patients.

Sharing data within a company tends to be tightly restricted because people feel like they have ownership of “their” data. Christian is pushing a model where people are seen as data stewards but not owners. This model gives them control over re-use rights for meta-data, which is very important. Another possible solution to encourage people to share more is to provide licensing rights for data scientists. Part of that process is to have the data scientists go through training with the clinical team to better understand what they can and can’t do with the data.

Use of AI/ML in the Short Term and in the Future
Short terms benefits of AI/ML tools are focused on operational efficiency (do things better). In clinical trials people are starting to shift to doing better things. Identifying sites and predicting failure rates in order to make trials more efficient. The most innovative work is being done in the research and scientific areas looking at whether cell lines are being productive, as one example.

Noisy Data
You don’t want large batches of data that are wrong misleading your model. However, when using models you are going to get noisy data so in some ways it’s good to train with noisy data. You need to look at the signal to noise ratio and ideally you want that signal to noise as high as possible, but you don’t want it higher than the real life data you are going to be using the model in.

Part 2

Executive Viewpoints Panel: Using AI Effectively from R&D to Patient Care – Critical Technical and People Issues with Mark Borowsky, Andrew A. Radin, and Niven R. Narain.

Background Information

twoXAR Pharmaceuticals was created six years ago. They are currently pursuing 18 different diseases with development candidates. 6 of those programs are licensed to other pharmaceutical companies and the rest twoXAR is developing.

Patients are still treated based on what is reimbursed. Niven wanted to focus on moving from genotype to phenotype and BERG takes a new approach using data-driven patient biology to drive hypothesis and in so doing works to democratize and desensitize biases. Simply stated, this approach takes math plus biology to drive new insights in medicine. Today BERG is a patient-driven biology company that drives biopharma assets by using AI tools to create more predictable and reproducible hypothesis. BERG’s assets include 3 late stage clinical assets, a number of oncoming phase II trials, along with a pipeline of over 25 preclinical assets.

Mark’s focus has always been on bringing the opportunity of quantitative methods to individual researchers. At NIBR Mark started with a mandate for building up a coordinated science team that would coordinate with distributed data scientists across the organization.

The Impact of AI/ML Tools in the Value Chain

Computational methods are a toolbox and you have to figure out what problems you are going to go after and how will you use those tools to construct solutions. twoXAR Pharmaceuticals is pursuing new ways or new ideas to address unmet medical needs from a biology perspective and then using computational screening technology to be able to move quickly to the wet lab to try out new chemistry against new targets. From an efficiency perspective this is extraordinary. The success rate of using this computational method with AI is also quite higher than traditional methods.

Along the supply chain from discovery to pre-market access, BERG is focused on how to pivot the in-silico insights to create more high-quality targets. Using these methods BERG has been able to move with a particular asset from an in-silico model to pre-IND enabling trials in 22 months, which is very fast compared to industry standards.

NIBR has seen a lot of promising results around target ID and lead optimization. Other results have been seen in lead discovery for some target classes and some progress has been made around mechanism of action. Other work is focused on seeing if they can predict how some molecules will fair in early stage proof of concept trials.

Partnerships Always Go Back to People

Partnerships come in two forms. One type is with a pharmaceutical company and the other partnerships twoXAR has entered into are with investors. At the end of the day, they want to see a clear path to the clinic. In partnership with pharma they are thinking about how they will meet an unmet medical need.

Partnerships always go back to people. Good partners need to have humility – they need to be humble and embrace the complexity of what they are trying to do and admit what they don’t yet understand. Good partners understand that they are in business to serve the patient. Good partners also care about value and are focused on creating value for the ecosystem they are part of.

Mark is looking for partners who have compatible therapeutic interests. The other piece they look for is a demonstration of real value and downstream work to ultimately get therapeutics to patients faster.

The Impact of AI on Speed and Efficiency

Patient trial selection has the potential to be speeded up using AI tools. Decision making in the discovery process can also be speeded up by using AI tools but it’s also equally important to understand what can’t be speeded up.

Time is efficiency and value endpoints. If you can have an enriched phase II trial that is speeded up using AI tools, that is a win. The clinical trial might not be faster, but the drug is launched in a more effective manner.

Repurposing is also an opportunity to use AI to speed up the development process. However, the real value of AI will also be about what new spaces AI/ML can unlock and that is the truly exciting aspect of this technology.

Final Thought

The fundamental question for everyone, whether you are from a large company or a small company, is how can we leverage this data to improve discoveries and cut down time to get therapies to patients

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 Biopharma Data Management Strategies to Effectively Use AI Tools, register for the upcoming interactive digital summit on September 15th. Speakers include thought leaders from: AbbVie, Novartis, Regeneron Pharmaceuticals, Inc., Unlearn.AI, The APANDEMIC Initiative, 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: ecutler@coreylanepartners.com.