Failing faster using AI cannot be considered a success – A round-up of TechBio UK 2023 conference


Bristows attended TechBio UK 2023 last month, the BioIndustry Association’s conference that showcases the latest developments from companies working at the interface between biotechnology and technology. It was an action-packed day, filled with sessions on a range of interesting topics from doing deals with big pharma and big tech to the use of AI and the future of genomics.

Here are our four key takeaways from this year’s conference:

1. Access to high fidelity data continues to be key: In the age of AI, it is unsurprising that high fidelity data is likely to underpin most TechBio products and services and a sizeable portion of a TechBio company’s value may be attributable to the value of the datasets it has access to. Building or securing access to rich proprietary datasets can be pivotal in justifying any valuation as well as gaining an edge on the competition. For those looking to the NHS as a source of data, there are other issues. While the treasure trove of data stored by the NHS is seen globally to have great potential, getting access to it is a complex and fractured process, with NHS Trusts storing data of variable quality independently in a range of disconnected systems. The UK needs to find a way to harness the NHS’ data potential if it wants to lead the way on the TechBio global stage.

2. TechBio companies should ensure their technology remains their focus: A popular business model for a TechBio company is to develop and use a platform that enables it to collaborate with biotech and pharma companies to fuel their drug discovery pipelines (for example, by enabling faster target discovery) while simultaneously using the platform internally to develop its own drug candidates. It can be very tempting for a TechBio company to turn its focus to developing these drug candidates, but it is crucial they do not neglect the golden goose, its platform, in the meantime. If it does, it risks customers moving to competitors, being left without a consistent revenue stream and with assets that are statistically unlikely to make it through clinical trials.

3. TechBio companies need to be adaptable with their contracting approach: It is well versed that a different approach to contracting is needed when big pharma negotiates with big tech companies. However, there are also some notable contrasts between the approach to TechBio companies concluding development agreements with big pharma versus the relatively smaller biotech companies:

· Deals with biotech companies tend to be shorter term agreements with more narrowly defined development goals agreed up front. This makes them much faster to negotiate but means TechBio companies are lacking longer term revenue streams necessary to ride out market volatility.

· While a deal with big pharma often provides a longer-term commitment spanning a wider range of development opportunities, the deals tend to move at a much slower pace with more stakeholders involved. Big pharma is also likely to want more control over the project. For example, the ability to step-in and take over if key milestones are missed. They may also insist on source code being placed in escrow to alleviate concerns around a small digital health tech company’s insolvency.

· What is true for both camps is that untangling and holding onto proprietary IP continues to be the main challenge for small TechBio companies.

4. AI is yet to have a truly transformational impact: There are a growing number of claims that AI will radically shorten the drug discovery and development timeline. So far, it appears that while AI has been highly effective when deployed in certain functions in the development process (for example, in identifying and optimising small molecules, and interpreting or finding patterns in patient data), this is only the start of truly transforming the process. While several promising drug candidates developed by leveraging AI have performed well in the lab and are progressing to early-stage clinical trials, it will be some time before we see if any of them succeed in humans. After all, reaching failure faster using AI cannot, in itself, be considered success. However, some are predicting that truly transformational impact will come within the next 5 years.