August 5, 2020
Biotechnology focus: developments in AI and machine learning
In this article, written for The Pharma Letter’s Expert View section, Mathys & Squire associate Amy Nick examines the latest developments in artificial intelligence and machine learning in the biotech sector.
Developments in artificial intelligence (AI) and machine learning (ML) are playing an increasingly influential role in the biotech space, driving the formation of new partnerships between the tech and healthcare industries.
While big pharma is increasingly seeking collaborations with AI startups, major tech players such as Google, IBM and Microsoft are taking steps into the biotech space. A 2019 survey of pharmaceutical and biotech professionals by ICON found that 80% of respondents were using, or planning to use, AI technologies.
AI has already begun to make an impact. Food and Drug Administration approvals of AI algorithms have increased exponentially over the past few years, and the AI healthcare market is predicted to reach $6.6 billion by 2021.
The potential value of AI is already apparent in areas such as diagnostics, with research suggesting that diagnostic algorithms can match the performance of human experts in detecting diseases from medical images. The use of AI in other areas is more speculative: no AI-designed drugs have been approved yet, and few have reached clinical trials.
This is in part because validating AI/ML predictions remains expensive and time consuming, especially where this requires synthesising entirely new compounds or conducting large-scale clinical trials. Meanwhile, the margin for error in healthcare is narrow, and even the most advanced machine learning models can only be as good as the datasets they are trained on. Potential investors need to be convinced that the rewards are worth the risk.
This will require a shift in approach to the creation and protection of IP. The patent system has long served as a mechanism for promoting and rewarding investment in innovation, but AI/ML innovation poses numerous challenges to this current system. As of yet, there is no consensus in the approach taken by national courts and intellectual property offices to resolving these, leading to uncertainty and inconsistency between jurisdictions.
One key issue raised by AI inventions is the nature of authorship and inventorship. In copyright law, it has long been debated whether a human creator is required for copyright to arise in a creative work.
This too remains unresolved. Some jurisdictions, such as the UK, award copyright for computer-generated works to the person responsible for facilitating creation of the work. In contrast, courts in the USA and Australia have denied protection to works with no substantive human input, although how these decisions will be applied to AI/ML remains to be seen.
AI and inventorship
Resolving the issues around copyright protection will be relevant for the protection of source code. However, AI is increasingly becoming capable of generating patentable output with diminishing human supervision. Here, inventorship is generally understood to reside with the person who developed the AI, which is seen as a ‘tool’ of the human inventor.
There are as yet no specific legal provisions addressing the notion of AI as an inventor, and most jurisdictions require the named inventor to be a natural person. Both the UKIPO and EPO recently rejected patent applications because the named inventor was an AI, despite acknowledging that the criteria for patentability were met.
The USPTO followed suit, arguing that US patent law limits inventorship to natural persons. Such refusals are unlikely to be the end of the issue. As unsupervised learning algorithms become more complex and their use more widespread, cases where human oversight over the final output are not sufficient to meet the legal criteria of human inventorship will bring this point back to the forefront.
Another question lies in the interpretation of patentability requirements. Computer programs are excluded from patentability in many jurisdictions on the basis that they can be protected by copyright, and similar restrictions apply to algorithms and mathematical models. Yet, it is not always clear how these patentability requirements should be understood in the context of
AI and ML. The EPO updated its Guidelines for Examination to include specific guidance on this for the first time in 2018, but the approach of many other patent systems remains unclear.
Broader questions also arise from the use of AI in biotech. Standards for inventiveness may need to be revised, as AI interprets and processes information in an entirely different way to a human inventor. Under current law, to obtain a patent an invention must not be obvious to a ‘person of skill’ in the relevant field on the basis of publicly available information. It is unclear how this should apply to AI-generated predictions.
AI-based approaches could conceivably identify drug candidates which are an obvious outcome of the application of AI, despite not being obvious to a human expert on the basis of the same information. The more commonplace AI becomes, the more difficult it may be to determine inventiveness exclusively by reference to a human inventor.
The IP strategy of companies in the biotech space will also need to evolve to meet new challenges. This is particularly true in the context of personalised medicine: as AI/ML-driven personalisation targets increasingly smaller patient populations, drug-makers may end up with treatments applicable to only a handful of patients, or even to a single person.
In such cases, traditional strategies protecting a specific composition of product or a particular treatment protocol are likely to be of little commercial value. Instead, companies will need to find ways of capturing value across all stages of the clinical development process.
In particular, protecting novel strategies for accelerating drug discovery, improving patient selection and enabling treatment optimisation, as well as innovative methods of data capture and the analytics tools underpinning them are likely to be increasingly important.
Legal, ethical and regulatory issues
The growing use of AI has given rise to a new set of legal, ethical and regulatory issues which must be addressed if innovation is to keep pace with technological progress.
A patent system able to adapt to these challenges will be key in allowing researchers to cooperate openly; without robust systems for protecting their IP, developers may choose to keep novel AI and ML processes trade secrets, depriving the research community of the opportunity to build on their progress.
Navigating these issues will be complex, requiring cooperation and discussion between the tech and pharmaceutical industries and the legal community, but this is vital in order for AI and ML to realise their full potential in healthcare.