The potential value of digital and analytics (DnA) to R&D in pharma is clear: leading companies are harnessing DnA’s power to reduce the time needed for early-stage drug discovery by between 20 and 50 percent and to cut early development costs by as much as 50 percent, for example. These are significant achievements in an industry that has long struggled to break even on its R&D investments, helping to bring better therapies to patients faster.
But such achievements are not widespread. While most companies are working to build their DnA capabilities, few have yet succeeded in embedding them deeply throughout the R&D organization to transform its innovation and performance.
Our experience working on transformations with companies around the globe suggests that successful DnA transformations follow a common approach. That approach entails setting an ambitious target for what a transformation might achieve that is in line with the company’s strategy, drawing up a plan that outlines the initiatives that fulfill that ambition, and making the organizational changes needed to ensure the company has the right talent and leaders to support the transformation.1
While most companies are working to build their DnA capabilities, few have yet succeeded in embedding them deeply throughout the R&D organization to transform its performance.
Some transformations flounder because they fail to get these basics right. But even companies with a solid transformation plan can fail at the next and often most difficult step of the transformation—its implementation. Some turn to partnerships with AI-enabled drug discovery companies for help rather than building their own capabilities. This can result in AI not being integrated into the day-to-day work of the R&D organization and therefore failing to take root.
Other companies have road maps for building some of the components required to support the initiatives—around modernizing IT infrastructure or setting up data governance, for example. But these tend to be developed separately with relatively few links made between them, and the precise goals and the value it brings to the business are often unclear. Work on each program can also be patchy, as developers, data scientists, and engineers can find themselves working on several initiatives simultaneously. As a result, it can take months or even years to deliver anything of value to R&D teams, and the results seldom seem to justify the resources and effort. For example, one company that had embarked on a costly two-year program to roll out an electronic lab notebook solution eventually discovered it was of limited use to both research and data scientists, as it missed key experiment information and had an overcomplicated data structure.
Still other companies find themselves in pilot purgatory—that is, conducting a large number of proof-of-concept initiatives that never become available at scale or deployed by end users: the people in the lab or those designing and running clinical programs.
This article describes a new approach to implementation that we have developed and that is being used by a handful of life sciences companies to overcome such pitfalls and drive impressive, speedy results in their R&D organizations and beyond. Solutions of real value to end users are delivered every 90 days, continuously building support for the transformation and thus helping to build scale. It is an implementation engine we call quarterly value release (QVR).
Quarterly value release: The concept
QVR rests upon six building blocks (Exhibit 1):
- A blueprint linked to value: a blueprint lays out the scientific or operational value of a particular initiative or use case as laid out in the transformation plan and the road map and backlog to develop a scalable product.
- Digital and analytics capabilities: the capability to develop analytics models and digital user interfaces that address the needs of primary users.
- Data architecture: the collection, structuring, standardization, and linking of common R&D data types, including preclinical and clinical data; the results from chemistry, manufacturing and controls; and data from relevant external sources.
- Technical infrastructure: the infrastructure required to build, host, and deploy data and analytics pipelines—a machine learning operations (MLOps) tech stack, for example.
- Talent and an agile operating model: cross-functional teams with capabilities such as computational chemistry/biology, DevOps (software development and IT operations), data engineering, and user-experience-driven design. Teams typically adopt agile working practices.
- Adoption and scaling: the ability to collaborate and co-create with end users to encourage adoption and support for the transformation, to track adoption, and to structurally embed changes in processes.
Many companies seeking a DnA transformation will have embarked on programs to establish these blocks, though they often concentrate primarily on the technical ones. What is different with QVR is the way in which the blocks are integrated with a view to rapid, value-driven implementation. Rather than treating them as separate, parallel programs, QVR brings them together through cross-functional teams that focus on a common, 90-day goal.2 That goal is to demonstrate a solution with “full stack” capabilities—that is, a solution that encompasses all six building blocks, is fully running in production, and can be operated by the end user. The solution will also deliver a significant improvement in a key scientific or operational measure, as detailed in the blueprint—an increase of the probability of success and speed, for example. In this way, what are typically horizontal programs become a series of bite-sized, integrated packages, each delivering on its blueprint (Exhibit 2).
This integration accelerates progress. The strong links between building blocks avoid the inevitable transactional costs incurred when each horizontal program has its own leader with his or her own priorities and incentives. In QVR, one product owner is in charge and responsible for alignment across all building blocks. Moreover, because there is constant communication between the different blocks, it becomes much easier to address problems that might arise or to spot opportunities to accelerate progress still further. And because end users are involved throughout the process, the value of the solution is assured and, hence, its adoption secured.
Another key element of QVR is the demonstration of the solution’s value at the end of the 90-day cycle. Transformation programs are typically reviewed by a steering committee by means of a presentation. But with QVR, it is the end users who demonstrate the solution live to R&D leaders. These demos not only keep the team focused on delivering value within the 90 days, but they also shape the entire transformation, as the end of a QVR marks an important decision point. Some initiatives might be completed in a single cycle; others might require several cycles, depending on their scope. Funding will always depend upon progress made to date, and R&D leaders could decide not to back further cycles but to prioritize other initiatives instead.
In this way, the implementation of the entire transformation is heavily informed by the value delivered every quarter, giving company leaders the opportunity to keep steering toward where the most value lies. One QVR team that was working on building a predictive translational model was able to demonstrate a correlation between candidate features and the biomarker response at the end of the first cycle. Nevertheless, departmental leaders decided to stop the program after three months as there wasn’t sufficient data to build a more robust model. In a different setup, that decision might have been made much later. It isn’t uncommon for companies to spend two years building a digital R&D platform to support some ten or more use cases, only to find that just two or three are useful to the wet-lab team—a costly waste of resources that could have been capped with more frequent reviews. Even the best-laid transformation plans will need reassessing as the transformation progresses.
How it works
To demonstrate the QVR concept, take an initiative to develop a closed-loop, high-throughput screening (HTS) system for small molecules—an initiative that is part of an overall transformation effort to expedite drug discovery timelines while increasing the probability of success of leads generated (Exhibit 3).
The QVR begins by preparing a blueprint for the first cycle, building block number 1. This sets the precise objective for cutting the time it takes to screen the compound library in silico by 30 percent. The blueprint also provides a product backlog for the next 90 days that is based on a clear understanding of how analytics could shape the screening process in a way that would be useful to research scientists and cheminformatics experts, and it defines the KPIs that will measure and track the cycle’s progress—in this case, the number of small molecules/compounds screened per month, the percentage of compounds with a “hit,” and the deviation between predicted results and experimental results.
Next, working in agile development cycles, the team builds and automates in silico screening algorithms for small molecules and builds a user interface for visualizing algorithm inputs and results (building block number 2, digital and analytics capabilities), while industrializing data engineering pipelines (building block number 3, data architecture). In the HTS example, this includes building a centralized data layer that collates experimental results and algorithm outputs in order to verify prediction accuracy and train the next generation of models. The team also enhances the technical infrastructure with new technical capabilities (building block number 4). Here, that might mean building a reusable software development kit and standardized, reusable data engineering pipelines, including the routine extraction, cleanup, and structuring of experimental data.
Simultaneously, training takes place to improve the agile working skills of the team, working in sprints and adhering to a product backlog that prioritizes value for the end user (building block number 5, the talent and operating model). This value focus paves the way for building block number 6, adoption and scaling. Because the company’s small molecule research scientists have been involved in building the algorithms, deciding the new screening protocols and processes that incorporate the outputs, and deciding on the KPIs, adoption is far more likely.
The QVR cycle ends with the research scientists (instead of the development team) demonstrating the solution and its value against the prioritized metrics to the heads of discovery or R&D to confirm their support. In this case, the solution is a machine learning algorithm for in silico screening automatically fed with experimental data that is embedded in a user interface and that expedites discovery timelines for small molecules (see sidebar, “Industrialization versus experimentation”), and scientists ran a screen for a new library using the novel algorithm to demonstrate its value.
The demonstration is also used to explain the proposed blueprint for the next cycle, which in this case is to add additional compound libraries and screening algorithms with a view to predicting and improving the synthesizability of the leads. Given the encouraging results to date in our example, the departmental leaders agree to fund the next QVR cycle.
Experience suggests a few principles that those embarking on a QVR approach to implementing a DnA transformation should keep front of mind, and some actions that will help keep them on course.
- Take a full-stack approach to the problem. In other words, do not focus only on the technical development components. The blueprint, capability building, and adoption and scaling are all key. The relative weight of each building block can vary: in a specific cycle, the team can, for example, spend more time on the data architecture; in a next cycle, the balance can shift toward adoption.
- The QVR concept is best run with at least three to five teams working in parallel to ensure a transformation takes hold in your R&D organization.
- Infuse the teams with a strong end-user mindset so that solutions take into account the challenges users face and deliver value. High-caliber product owners capable of blueprinting, prioritizing, and engaging end users are a must.
- Don’t be tempted to skip the demonstration of a full-stack solution by end users at the end of each QVR cycle. You need to be determined to deliver value, and the demos act as a strong incentive to reach the targets laid out in the blueprint.
- At the end of each cycle, consider whether the overall plan needs adjusting. The most successful transformations consistently review progress to date, closing down initiatives that fail to deliver value and backing new opportunities uncovered.
- Agree on a decision-making structure for product development, with the product owner in overall charge. And keep a decision log for accountability. This will help overcome the issues that arise in cross-functional teams whose members may have different priorities or risk concerns.
By merging the various programs of a transformation in a series of QVRs, life sciences companies can overcome the typical challenges they face when implementing a DnA transformation in R&D and beyond, while shooting for an ambitious end goal. Each QVR delivers value to end users in just a few months and encourages adoption and deployment of the solution at scale. Multiply the effect with several teams using the same approach to develop a series of initiatives, and DnA capabilities quickly become embedded across the entire R&D organization. This can only help to accelerate drug development to the benefit of patients.