The new high-speed drug discovery path using AI: Where soft IP meets hard IP to form a dynamic alliance
by Dr Eileen McMorrow

A fusion of technologies is happening which promises to revolutionise the drug discovery process, and Artificial Intelligence (AI) is a key part of this. Various types of Machine Learning (ML) demonstrate what AI is capable of, and the use of such tools is likely to greatly shorten the drug discovery path. Vast amounts of data relating to research papers, small molecules, protein structures, genetics, and animal and human clinical trials can be analysed using ML and other AI capabilities.

Collaborations exist between businesses that create algorithms and software for accelerating drug candidate selection, and those that specialise in small molecules, protein chemistry or other clinical or biological aspects of drug design. These include Atomwise with Merck and Abbvie, Exscientia with Sanofi, Insilico with GSK and Numerate with Takeda.

BIA member Benevolent AI uses machine learning to identify potential drugs and effective targets for specific diseases. They recently bought Proximagen’s drug discovery unit to combine with their AI capability.

So what kinds of IP are these collaborations expected to produce? A mixture of IP may arise including hard IP in the form of patentable inventions and soft IP in the form of copyright, database rights, and rights protecting know-how and confidential information, and rights in designs, trade secrets and trademarks.

What IP can exist in data?

• Database Rights

The 1996 EU Database Directive required all EU member states to recognise database rights. This was implemented in UK law by the Copyright and Rights in Databases Regulations 1997. Database rights automatically protect collections of works, data or materials arranged in a systematic way that are individually accessible where a substantial investment has been made in obtaining, verifying or presenting the contents. They last for up to 15 years. Any substantial modification of the database creates a new database right. Recent healthcare cases demonstrate their importance.^1,2

• Copyright
  
  

Databases are also protected by copyright if there has been sufficient intellectual creation in the selection or arrangement of the contents.

Is there potential for both Hard and Soft IP in the algorithms and associated software involved in AI?

Patents

It has traditionally been easier to patent software in the US than in the EU, but following the Alice decision³ it has become more difficult to satisfy subject matter eligibility and requires a two-step test.

The United States Patent and Trademark Office (USPTO) has however granted a large number of patents in the field of AI under the Cooperative Patent Classification (CPC) – class 706: Data Processing –Artificial Intelligence.

Under the European Patent Convention (EPC), a computer program claimed "as such" is not patentable. However the EPO does grant patents for inventions in many fields of technology in which computer programs make a “technical contribution”. This technical contribution is often difficult to demonstrate. Recent decisions from the UK courts have provided further guidelines for UK patents.⁴

Trade Secrets

AI is a fast moving area, and companies may prefer to avoid the prolonged software patent application procedures and the disclosure of the underlying algorithms. Trade secret laws may provide another means of protection if the algorithms satisfy the definition of trade secrets.

The EU Trade Secrets Directive 2016 is due to be implemented in UK law in June as the Trade Secrets Regulations 2018. The US has the Federal Trade Secrets Act 2016 and most individual states have the Uniform Trade Secrets Act.

Copyright

Most software businesses rely on copyright to protect their IP. Software is treated as a literary work under UK copyright laws and is protected by copyright under the laws of most countries. Copyright protects against literal copying of the source or object code. To a more limited extent it can also protect the original architecture, design and functionality of a computer program, but copyright would not protect an algorithm as such.

How can the IP in the resulting drugs be protected?

Patents are likely to continue to be extremely important for those drugs which are discovered utilising AI as part of the discovery process.

Soft IP such as rights protecting databases, know-how and confidential information may be associated with the patents and when the drugs are marketed trade mark protection will be important.

Drug delivery devices may also be created using AI, where registered or unregistered design rights may be applicable.

Who will own the IP generated from AI enabled drug discovery?

The use of AI in drug development is likely to pose some interesting legal questions. What is the best way to protect and exploit those rights? Who owns the rights in computer-generated inventions? Attention will need to be paid at the beginning of any collaboration regarding access to background IP. Carefully drafted written agreements will be needed to clarify who will own the foreground IP generated along the AI drug discovery path and what licensing arrangements are required.

It will be interesting to see if new types of IP rights evolve as the AI drug discovery industry develops.