Representatives from start-up development company Wheeler Bio and its digitalization partners, DataHow and Synthace, discuss how their collaboration will help Wheeler gain deeper insights into its processes more quickly and accelerate the time to IND for its customers. The digitalization solutions under development will also enable Wheeler to share customer data in real time, helping to build trust and strengthen the partnerships it has with its customers.
David Alvaro (DA): How would you characterize the current state of the industry with regard to the incorporating of automation and artificial intelligence (AI) into process development?
Jesse McCool (Wheeler Bio): The biopharmaceutical manufacturing industry is behind other industries in terms of adopting digitalization tools. However, we have seen in the last several years a wave of “Industry 4.0” or “Pharma 4.0” success stories in large pharma that will drive further adoption of digitalization across multiple industry segments. With the latest high-throughput bioprocess development equipment and PAT (process analytical technology) instrumentation that make up the core of Wheeler Bio’s development capabilities, we felt it was important to incorporate digitalization strategies up front in the conceptualization phase of our services platform. We looked for solutions that would help our scientists generate more data with each experiment while studying more parameters and interactions in an easy to setup DoE (design-of-experiments) format. We looked for powerful predictive modeling tools that can save significant time in process development. We also looked for solutions that improve the linkage between and among different pieces of equipment, allowing our scientist to build automated lab workflows. And most importantly, we wanted to be in a good position to safely share data in real time with our customers and collaborators. We evaluated the options and are excited to move forward in partnership with two leading providers that are offering all these solutions now: Synthace and DataHow. The core of our mission is to increase access to manufacturing technologies by making it easier for customers to tech transfer and reach their clinical milestones. Digitalization of an open-source manufacturing process is how Wheeler is going to achieve this.
Michael Sokolov (DataHow): Although companies would like to be able to perform big data analysis, the biopharma industry is not yet truly data rich. A lot of labor is required to generate each data point, so it is necessary to survive in an environment that is low in data but high in uncertainty. Informed decision-making is crucial, because the ultimate result of the overall processing work is the creation of a pharmaceutical product that is very consistent in terms of its quality. The engineering team thus faces a considerable challenge to deliver consistent quality in what is often a very short time.
Today, the underlying hardware technology has been optimized, which leaves digitalization as the last approach to further improving protocols. Contract development and manufacturing organizations (CDMOs) are ideally positioned to benefit from digitalization, because they work with multiple molecules from different clients. CDMOs also have extensive internal databases that offer the potential to establish process-to-product relationships and enable them to leverage all of the knowledge created in platform processes.
These conditions are effective for data science. It is possible to use these data to create predictive models. In addition, as data science advances toward knowledge science — taking generalized data and converting it into general process understanding — such tools can be used to support decisions on the basis of similarities to previous experiences.
Of course, this is more complex than just relating what has been done before to incoming activities. Human beings are cognitively limited in what we can accomplish, which is why we need tools that enable consideration of multiple perspectives; they allow us to not only think and make decisions but to loop that information back into the system.
Looking forward, we want to advance into a biopharmaceutical environment where humans are supported by many such digital solutions. These technologies will not replace humans, but they should do a lot of the jobs that humans cannot do well, thus allowing the creation of better processes — ideally in less time.
These tools will give a company like Wheeler Bio a very strong competitive advantage, because it will allow them to achieve similar capacity as their larger competitors. For the industry as a whole, there will be a movement away from the current basic linear model to one that supports the natural, nonlinear nature of processes and their interrelationships.
Ultimately, the utilization of these digital tools over the coming years and across many stakeholders in the industry will allow us to raise the standard of how such drugs are produced and to enable the development of completely new drugs. For instance, for personalized medicines, such as cell and gene therapies, there are currently even less data and more complexity, but decisions still need to be made to support the goal of delivering these treatments more quickly to patients. In such an environment, these tools not only improve quality but fundamentally enable such therapies to reach the market.
Markus Gershater (Synthace): The biopharma industry has to make use of 21st century digital tools and automation. We are trying to engineer the most complex things in the known universe, and we’re doing it mostly with whatever springs to mind or is within arm’s reach — essentially e-copies of Excel. That simply isn’t good enough. To really penetrate into the heart of what’s going on with their systems and achieve deeper understanding than is currently possible, researchers need sophisticated modeling, predictive modeling, digital twins, and AI augmenting their abilities.
We all recognize that the only way we can build this is on a foundation of exceptional data — not just the primary data, but the metadata of how those data were produced. Metadata provides critical information about context — the story that led to the result. If we digitize the way that experiments are designed and run and make sure that we’re capturing all the data that result from those experiments, we can fully and completely map exactly what happened in the lab to produce the data.
This means we can provide results to scientists as full data sets with all of the context baked in; they’ll know everything about their data and exactly how those data were produced. This is where we need to get to, because both biology and the experiments we run with it are exceptionally complex. We need to ensure that we really understand those experiments in full and have that digital representation of the experiments we’re running.
We also need to consider how our relationship with the lab will change once everything is digitized and in the cloud. We can design experiments from wherever we are and run them in the lab. Clients of CDMOs like Wheeler Bio who want to know what’s happening with their experiments could log into the same system and see the data. Digitalization tools can benefit many human aspects within the biopharmaceutical industry, particularly around communication and collaboration.
Guy Levy-Yurista (Synthace): I came into biopharma from big data. It was in that space where I first started focusing on machine learning and AI. I believe that the 2020s will be the life sciences decade, including not just biopharma but also agritech, food tech, climate tech, and so on. Leveraging biology and the life sciences sits at the heart of solving problems in these sectors, and the complexity of biology demands that we go beyond the limitations of human ability when working with spreadsheets, or even pencil and paper. We need digitalization to enable the hyper-dimensionalization of data — this is how we will integrate many small, disparate data sets to accelerate the rate of progress within the life sciences.
DA: What was the specific impetus that led to the partnership between Wheeler and Synthace, and what does each organization bring to the collaboration?
Markus Gershater (Synthace): Jesse and I have known each other for ages. We’ve been passionate advocates of DoE and multidimensional experimentation. When we first met, it was pretty rare to come across others who were such passionate advocates of this approach to development. At Synthace, we were constantly attending conferences to try to understand the most current approaches to managing this concept. That is where I met Jesse and other Wheeler folks.
Christa Short (Wheeler Bio): Over the last 10 years, many of the Wheeler team members have been involved in multiple process validations at several companies –– especially around process design (or process characterization), process performance qualifications (PPQ), and continuous process improvement projects. The scopes of work tied to process characterization programs require significant resources and are incredibly important to set the stage for a successful PPQ. These programs typically take advantage of DoE to explore many different input parameters and parameter interactions. However, it takes a lot of resources, skill sets, extensive experimentation, and data analysis to perform and analyze complex DoEs, and as a consequence a lot of process development involves one-factor-at-a-time (OFAT) experiments, especially when there isn’t a lot of time or budget for process development. Today, with the right tools, it is more straightforward for process scientists to use DoE even in early development. This is what we are aiming to do here with this collaboration: to give our scientists access to tools that enable more data points and process understanding.
Yuk Chun Chiu (Wheeler Bio): As one of the founding Wheeler leadership team members with large pharma experience, I am excited by the level of engagement that our process development team has shown in seeking out and taking advantage of digitalization tools. They share the culture of the big pharma R&D groups. These software products allow us to efficiently apply QbD (quality by design) principles to our mAb platform development strategy. However, one challenge that CDMOs face in developing platforms in a fee-for-service model is that the process, products, and data do not typically belong to them. The confidentiality surrounding customer projects is a sacred commitment in the CDMO world. Wheeler is tackling that challenge by using a library of in-house molecules — sequences shared by our partners and from the public domain — to generate the data points and, ultimately, our own bioprocessing model. Christa’s team is currently working with three molecules at five different scales and multiple platforms for upstream and downstream processing. Because they are in-house molecules, we can be infinitely more transparent and collaborative with our platforms once complete and ultimately create more value for our customers with open-source access.
Christa Short (Wheeler Bio): Also, these software products are critical for us to meet our own business goals and have a compelling platform available to customers in a timely manner. Our goal is to build a robust, scalable mAb platform in less than 12 months. We also need to consider the needs of the bispecific molecules that are in the pipeline. All of Wheeler’s clients are moving from late discovery into development for the first time and need to have timely access to manufacturing to scale up material very quickly and as cost effectively as possible. The more insight we can gather about our platform now, the more technical risk we can eliminate for our customers in the future, which makes us a more competitive CDMO.
Our platform features next-generation cell lines and state-of-the-art bioprocessing equipment, PAT, and powerful analytics, including the Ambr® 250 system from Sartorius, which comprises 24 fully featured single-use 100–250 mL mini-bioreactors. Our scientists are generating a lot of data and require efficient means by which to capture, organize, and analyze. With Synthace and DataHow, we can focus on data and make faster decisions. With this volume of data, and the ability to analyze it efficiently, we will be able to learn and understand our processes more deeply and more quickly, which gets back to reducing risk. We also hope that we might eventually be able to reduce process run times for mammalian cell culture from 14 days down to 10 days or less without sacrificing insights.
David Schmidt (Wheeler Bio): From an analytics standpoint, the new software will also help us drive down errors commonly associated with high volume of samples and increasing number of analysts. More analysts naturally means more operator errors. The digitalization software combined with high-throughput automation allows us to structure the data around the method and drastically increase the number of samples while reducing the number of operators, thus reducing the risk of errors. That means we can create high-quality and better-structured data that feed into some of our predictive modeling and other type of analyses that we’re doing.
DA: How did DataHow become a partner in executing this vision?
Jesse McCool (Wheeler Bio): We met Michael at a conference several years ago in Berlin. We were impressed with him and his team of data scientists and how he was positioning his company to synergize with pharma companies. Early discussions led to multiple meetings, site visits, and pilot programs and finally now they have a unique product that is tuned and ready for Wheeler data. DataHow enables easy access to hardcore bioprocess modelling expertise so we can be more capital efficient without sacrificing the quality of data science that’s needed to develop a scalable, robust biologics platform. DataHow is a leader in bioprocess modelling trying to help the industry grapple with an inherent scarcity of data points –– because running drug substance batches is very expensive. We are thrilled to have them on our tech development roadmap.
Michael Sokolov (DataHow): We have to bridge the fact that we do not truly have a lot of usable data in biopharma. We know the underlying processes, and as engineers we recognize that digitalization can tell us what we have not yet learned — how we can go beyond the cognitive capacities of human beings. The overall goal of data science tools is to help engineers and scientists make better decisions, faster. A solution that augments the current workflow in biopharma by using AI in the background to support decision-making is in high demand.
The key is that data science isn’t just for data scientists. It should be available for any end user, for any scientist, for any stakeholder who is somehow operating along this workflow of bioprocesses. That also means that the solution must be in a framework that is user-friendly for all of the people involved. It must be automated and make suggestions in the background about the best model or tool and then present the final basis for use in decision-making.
To become a standard, a tool must not only be trustworthy but also educational for the community of end users if it is to be considered, similar to how bioreactors and chromatography columns are viewed. Without a high level of buy-in from users, it is very difficult to make data science work in such an easy way. As data and digital solution providers, we have a very important educational commitment to ensuring that data science drives across a broader community.
We listen closely to the feedback of our end users and ultimately create solutions together. For clients like Wheeler Bio, that means developing customized solutions centered around their workflows and the standard decisions they must make.
DA: What real, practical benefits do you expect to realize initially, and how do you see that evolving over time as models continue to optimize?
Markus Gershater (Synthace): Short-term benefits are reduced data management and easier data analysis but with better insight into what the data actually mean — it’s a lot less pain and a lot more insights. That insight then compounds over time. We are not talking about a minor change, either. Once the broader industry succeeds at digitalization, it will be a truly transformative shift from where we are today.
David Schmidt (Wheeler Bio): From the perspective of a service provider, this technology will be transformative for the customer experience and the quality of service. It will allow our early clinical programs to move faster and faster, which is a key value driver influencing customer buying decisions. Thus, especially when competing for the venture-backed biotech business, this will be a key competitive advantage for Wheeler Bio. We anticipate that this collaboration will enable Wheeler Bio to reduce the time to an IND by at least three to six months, and hopefully six to nine months. That will be achieved by a combination of digitalization and a novel business approach that involves integration with contract research organizations (CROs) who are upstream of Wheeler.
We also have an NCI-designated cancer center across the street and look forward to connecting our customers with those resources to see what synergies can develop.
Guy Levy-Yurista (Synthace): From Synthace’s perspective, the benefits include an order-of-magnitude improvement in the time it takes to get to results and the quality of the science produced, as well as a significant reduction in cost. A year’s worth of science can be crunched into one month, during which time drug molecules can be produced and analyzed with science that is 10 times better than what would be possible without digitalization, hyper-dimensionality, and the use of AI and machine learning for modeling and complex data analysis.
Customer satisfaction is also a huge benefit. That is why we were performing initial experiments in our own labs until Wheeler Bio got their equipment installed and operational. We wanted to enable them, once they were ready, to hit the ground running. We are also continuously challenging ourselves to become better, because we know that we are helping others to create a better future for humanity.
Michael Sokolov (DataHow): I think that, when we look into the future, we first need to distinguish what is really possible today in terms of digitization, which is moving away from paper toward digital, and to recognize that this is just the first step. Going forward, digitalization will involve improving current business processes in terms of cost reduction, time savings, and performance gains. Ultimately, digital transformation will lead to completely changed business models — for CDMOs, certainly.
There is tremendous competitive advantage for CDMOs that can move faster in this direction than others, and that advantage can be leveraged in many directions. It will involve the creation of huge databases — knowledge bases to be exact — that provide the unique ability to manufacture completely new drugs that are not off-the-shelf and clearly out-of-the-box. It will also provide the ability to operate in high-throughput mode, smoothly.
Looking further out to what is coming next — I believe that, in five years, we will be deploying quantum computing. I’ve already instructed Markus and his team to start looking at how we can leverage quantum computing into true multi-dimensionality and analysis of complex biological systems. It is still very early days, but we are already looking at bioanalysis from that perspective.
DA: It strikes me that, when you take a bunch of current bottlenecks and optimize them by orders of magnitude, new pain points or bottlenecks that aren’t apparent or relevant right now might be revealed. Have you given any thought to what the next challenges might be?
Markus Gershater (Synthace): I don’t necessarily see us uncovering another big bottleneck like this. There will be others, but nothing so major as one that allows the whole biopharma industry to move so much faster than it has done before. When we were a process development company ourselves, the hardware was available to automate experiments, which made things easier to implement. We found that our software enables people to perform those automated experiments more effectively. Indeed, we shifted the bottlenecks to the data and to the understanding of those data. Overcoming this bottleneck provides access to insights into the biology of biopharmaceutical processes.
Jesse McCool (Wheeler Bio): Subject matter expertise is one of the least scalable resources in any business, but these solutions will change that reality. We think that new bottlenecks will be there for new modalities with much less foundational, commoditized process science; however, we view this data science initiative to be very rinse-and-repeatable. Although we’re laser focused on therapeutic antibodies right now, our CRO partners are continuously innovating around new modalities. Once we’ve applied the solutions from Synthace and DataHow to our laboratory workflows for mAbs, we can shift gears to other modalities.
Markus Gershater (Synthace): I just love the idea that we can use these technologies for ever-more ambitious goals. Yes, faster. Yes, with more certainty. Those are clear business drivers for what we need in this space. But for me, as a scientist, the reason I first got into all of this — and I think the reason that we probably all got into this in the first place — is that the work is about breaking down barriers and solving ever-more challenging problems.
Michael Sokolov (DataHow): I recently finished an MBA, and in my thesis I collaborated with many decision-makers across pharma to identify the key inhibitors to digital transformation. Two clear factors became apparent. One is the need to meet extensive regulatory requirements in order to get new technologies validated. Any new technologies, regardless of the advantage they provide, still need to go through a complex validation process. Hopefully, when regulatory authorities see the tremendous benefits, they will be more incentivized to encourage their use.
The second is that digitalization technologies are available, although some are clearly more mature than others. Most companies have some budget set aside for digitalization. There is a real need for education and training about how all of these digital technologies must be delivered in a very tangible and user-friendly way. Solution providers need to develop digital tools that work smoothly, so that new users will be more willing to accept them and not view them as nice but as necessary for improving major workflows. It is therefore very important that we ensure that digitalization becomes a commodity.
DA: From the points of view of Synthace and DataHow, how unique is this project with Wheeler Bio in comparison to other work you are doing?
Guy Levy-Yurista (Synthace): Jesse and the Wheeler team are really at the forefront, pushing the cutting edge of where things should be and where things are going. That is why there is so much alignment; we are all pursuing the future. In that respect, we are running as fast as we can, right by Wheeler’s side. DataHow also makes perfect sense to us. Again, it’s all part of the same vision: this notion of dropping in all the data, making sure that we make use of it all, and taking the best advantage of it.
Michael Sokolov (DataHow): DataHow started as a spinoff from a university, and we initially would convince scientists of the advantages of our technology and then hope they brought management onboard so we could get a budget for at least feasibility studies. With Wheeler, the CEO was looking for this type of solution, and so we have support from the top. It is a privileged situation for us to have a lot of buy-in based on trust and collaboration to see how we can then develop the overall approach.
There are two types of customers — those with a lot of trust in their data and what can be done with them, and those with little data but the expectation that much can still be done. Wheeler is a good example of the latter case. With a strong belief that digitalization can support decision-making and all of the stakeholders involved, there is a lot of positivity around using the technology and providing feedback, which makes the collaboration very, very fruitful.
Lance Johnson (Wheeler Bio): We think you’ll see the fruits of the labor pretty quickly with customer feedback. CDMOs serve customers, and those customers serve patients. Generally, CDMOs have a bit of a negative reputation for reliance on proprietary platforms and technology-anchored platforms that makes it hard for customers to get data and transfer projects from site to site, which drives up some of the costs of making pharmaceuticals.
The types of tools we are developing through this collaboration will have a big impact on the customer experience. We are building a CDMO service that puts customers at ease with a culture of transparency and partnering. They have access not only to our scientists and their knowledge and expertise but also to their data. When we share information — when we livestream data for clients — they are astounded. That type of service is not something they are typically offered. Sharing data in real time gives both parties the ability to view trends and collaborate to react accordingly to head off potential issues.
That’s our short-term goal. By doing so, we think we will have a real impact on the reputation of CDMOs in this industry. From an IT perspective, it is not overly difficult to deliver this level of service. You must break down the industry’s legacy paradigms and barriers, which is easy to do because the underlying concern is outdated. It takes stepping outside the normal conservative CDMO box and recognizing that there isn’t any greater risk to sharing client data with the client. Doing so elevates the level of trust between the CDMO and the client to the point where the parties become true partners, working toward the success of the process and product. What Wheeler Bio is doing by taking these data and putting them into a format that can be shared with clients and that they can easily understand is unique and rare and is a service very much desired by customers across the biopharmaceutical industry. It’s an exciting time!