Patenting materials informatics inventions in Europe
The history of science is littered with seminal discoveries that came about through serendipity. An example is the discovery of the sweetener aspartame; in 1965, James Schlatter was researching compounds to treat gastric ulcers when, licking his finger to pick up a piece of paper, he noticed a strong, sweet taste which he realised came from the compound – aspartame – he had just made.
An interesting thought experiment is to ask if aspartame could have been invented if James Schlatter had taken a less relaxed approach to laboratory health and safety. Could a researcher setting out on a research program to find a safe, low-calorie sweetener have discovered aspartame deliberately rather than by chance?
For such a researcher, plotting a path from known sweeteners to aspartame would have involved addressing numerous complex factors which would have made it difficult to accurately predict the properties of possible candidate compounds and so a lot of experimental trial and error would have been necessary. However, researchers are increasingly making use of informatics, which draws on a range of disciplines, including statistical analysis and machine learning, and uses existing data to shorten the amount of time and to reduce the degree of risk involved in developing new materials.
The use of informatics-based methods to identify new materials raises questions as to how innovations in this area can be protected:
- Can the materials informatics platform be protected?
- Can materials identified by the platform be protected?
In this article, we consider these questions with particular reference to European patent law.
Patent filing statistics shows that patenting in the field of materials informatics is beginning to grow.
International Patent Classification (IPC) code G16C 60/00 is concerned specifically with materials informatics:
G16C 60/00: Computational materials science, i.e. ICT specially adapted for investigating the physical or chemical properties of materials or phenomena associated with their design, synthesis, processing, characterisation or utilisation.
PCT publications in this field show a sharp upward trend over the last 5 years, starting from a very low base (Figure 1).
Looking at filings per jurisdiction, the list is dominated by China (2519 publications), followed by the USA (249), the EPO (157), Japan (153) and South Korea (149) (Figure 2).
This clearly points to a strong interest in obtaining patent protection for materials informatics inventions in Europe.
Patenting materials informatics platforms at the European Patent Office
One of the enabling technologies used in materials informatics is machine learning. In this section we consider whether a machine learning method for identifying a material having desired characteristics is patentable – for example, a method for identifying mechanical properties of an alloy in the field of materials science, or ability of a compound to act as an agonist or an antagonist for a certain receptor in the field of pharmaceuticals.
The most authoritative decision on patentability of computer-implemented simulations at the EPO is G 1/19 of the Enlarged Boards of Appeal of the EPO. The patent application in question in this case was drafted, prosecuted and appealed by co-author Dr Pawel Piotrowicz.
G 1/19 set out a coherent way of assessing claims to a computer-implemented invention including supporting the COMVIK approach to the assessment of inventive step (as set out in T 641/00) and confirming that a claim should be technical over the whole of its scope.
The key issue when considering any form of computer-implemented process (whether it be a simulation or one using a trained neural network) is whether the process solves a technical problem by producing a technical effect that goes beyond the implementation of the process on a computer.
G 1/19 considered how and when technical effects or technical interactions can occur in the context of a computer-implemented process. In particular, a technical input can consist of a measurement, and a technical output can exist as a control signal for controlling a machine.
Although there are only a handful of examples of European patents granted since G 1/19 was issued, they provide helpful insights as to how measurements or interaction with physical plant or machinery can lead to European patents being granted in the field of material informatics.
For instance, EP 3 809 118 B1 is directed to a computer-implemented method for predicting a property value of interest of a material. The granted claim recites the step of providing a population of infrared spectra of samples, in other words, measurements. The spectra are used to generate a prediction function and the property value is predicted using the prediction function.
Thus, a platform which uses real-world measurements and machine learning to help identify suitable candidate materials might be patentable at the EPO.
In another example, EP4 038 620 B1 is directed to a method for generating a thermoplastic composition for mechanically and/or thermally loaded component parts. The claim recites a computer system which has access to a database of known compositions and which is connected to a facility for producing existing compositions. The claimed method recites using known compositions stored in the database to train a neural network. The computer system identifies trial compositions using the neural network and then uses the real-world facility to produce and test the trial compositions. The test results are used to train the neural network. This can help lead to automatic, more reliable identification of candidates.
Thus, a platform which involves identifying potential candidates and then testing them, especially if this results in feedback, would also seem to be protectable at the EPO.
Given the rapidly increasing use of simulations and machine learning methods across a wide range of industries, we can expect EPO case law on patentability of material informatics platforms to develop in the coming years.
Patenting products identified by materials informatics
A product identified using materials informatics and then produced in the “real world” can be patented if it is new and inventive.
However, if a known platform is used to identify a product having desirable properties, then the inventiveness of the product may be in question. For example, if a known model for alloys is used to identify a new alloy having high tensile strength and if the new alloy is manufactured and is shown to have the high tensile strength predicted by the model then a claim to the new alloy might be objected to as lacking inventive step on the basis that there is nothing inventive in making an alloy which a known model has predicted would have high tensile strength.
This points to a future in which obtaining patent protection for a materials informatics platform might be at least as important as patents for materials that were identified, in whole or in part, using such a platform.
The growing use of materials informatics means that it is becoming increasingly important to obtain protection not only for the products identified using materials informatics platforms, but also the platform itself.