Our experiences with XOps in life sciences

The life sciences industry has long been driven by the need to develop and exploit solutions as quickly as possible, leading to it being very much at the forefront of technological innovation. The COVID-19 pandemic has further highlighted the need for the ability to adapt quickly – in diagnostics, drug discovery, clinical trials, and in scaling up manufacturing faster than ever before.

Technological innovation within life sciences

XOps is about the smooth delivery of technological innovations into sustainable and flexible production environments.  Some examples of technological innovation in life sciences are:

• Graph technologies in drug discovery. Pharmaceutical companies have been able to assemble biomedical knowledge graphs out of the known links between diseases, treatment drugs, and proteins and genes. These graphs make it possible to infer unknown node properties and links between entities that were previously not obvious. This in turns means that it is possible to find existing molecules that can treat different diseases than they were originally designed for.
• Combining deep learning and graph databases in drug discovery.  The discovery in 2019 of a new broad-spectrum antibiotic, Halicin, was made possible using deep learning models to predict the properties of new molecules.  These were trained on graph databases holding molecule properties, such as toxicity, binding energy, & antibiotic response. Halicin had previously been researched for treatment of diabetes and other uses, so much of the human safety research had already been conducted.
• Exploring the use of quantum technologies for modelling molecules. While quantum technologies are still undoubtedly in their infancy, there are some very exciting potential use cases that pharmaceutical companies are currently exploring with a view to scaling once the technology matures. Quantum circuits can be used to model small molecules or fragments of larger molecules, to predict certain quantum features of these molecules, for example covalent warhead reactivity. If you are interested in quantum in Life Sciences, do read some other blogs we have written:
  • https://www.capgemini.com/insights/expert-perspectives/exploring-the-value-and-limitations-of-quantum-inspired-solutions-for-life-sciences-organizations/
    
  • https://www.capgemini.com/insights/expert-perspectives/avoiding-quantum-pocitis-five-best-practices-life-sciences-organizations-can-embrace-at-the-outset-of-every-quantum-computing-proof-of-conceptstrongaa-href_msocom_1acr1anbspstrongstrong/

However, as the amount of data flowing into R&D teams continues to rise, there’s a risk of potential discoveries such as these going unseen. Technological solutions like AI are very helpful, but they’re not the whole solution. For that, we need to look at the foundations of the way R&D teams are organized.

Building an XOps organization

At Hybrid Intelligence, we provide staff with experience encompassing several specializations. We hire staff with scientific academic backgrounds. We then train them in important disciplines, such as manipulating and analyzing data, modeling, software engineering, and architecture. Many staff come with foundational aspects from academia, however we professionalize and imbue best practices, and we encourage staff to share their experiences and skills with their colleagues.

This approach to staff development forms the foundation for our XOps organization. We can also build an XOps team using a collection of specialists with a strong culture of collaboration and communication. When we form our XOps teams, we ensure to balance the team skills appropriately depending on the solution. We will include a mix of the relevant domain and technical experts.

Our experience in life sciences

We recently established an XOps team in one of our clients’ drug discovery organizations, to support a portfolio of innovative tools that have certain specialist purposes. Uses ranged from user-friendly ways to run models on an HPC used to find new proteins with similar attributes to others, to an automated way to identify systematic errors in lab results. We are working with various machine learning models, a graph database, a data pipeline step to ingest from other databases, and a lot of software with very specific UX needs.

Our XOps team is made up of several data engineers, software developers, some HPC specialists, and a small number of data scientists. We also chose team members based on their experience in biochemistry. The team is coordinated through an XOps manager, who has broad experience in all the areas mentioned previously.

When working, we involve representatives from each of the main functional areas that make up our team in all discussions. We find that we can not only provide a much better service, but solution reliability is increased, time to resolution of issues is lowered, and we generate more ideas as a team as to how to improve the tools. These have impacts on our end client that help improves the efficiency and effectiveness of their work.

I have also experienced the opposite – where there was not an XOps team in place, but in fact many disparate teams handling the maintenance of each component of the solution that aligned to their functional area. In these instances, I observed shirking of responsibilities, finger pointing regarding issues, and an inability to get tasks done quickly.

Conclusion

The situations I have outlined are currently bringing benefits to my clients. In the future, as emerging technologies mature and converge, solutions will grow increasingly complex. If XOps is not necessary for your team right now, anyone adequately managing the wave of new innovation will find it increasingly important in the future.