Data collection systems have only increased in recent years. Data alone has no value. But if processed in the right way, it is the key to optimising and improving any process. People do not have the capacity to process the immense amount of data that is collected today. Digitisation of data processing is the key to give value to data. And among the many tools that are already being used to do this, we find artificial intelligence.
An artificial intelligence is a system capable of acting autonomously by emulating a human behaviour. Closely related to artificial intelligence is machine learning. A machine learning model is able to learn from data, and create rules from this observation. These systems change the traditional logic applied to systems.
Traditionally, a system would analyse data based on rules predetermined by a human to give an answer. With the advent of artificial intelligence and machine learning, a system is now exposed to a set of data and the response to that data analysis. And it is the system itself that creates the rules that allow the system to respond. In this way, the more the system is “trained”, the more finely tuned the rules will be, and therefore, the better the decision making will be.
This technology has multiple possible applications. Its potential is enormous.
The pharmaceutical industry must also seize this opportunity. And it is already doing so. Perhaps one of the most remarkable milestones that artificial intelligence has provided specifically for the biopharmaceutical and biotechnology sector is the ability to predict the structure of proteins.
Until a few years ago, the structure of a protein had to be studied experimentally. (It is worth noting that the structure of a protein is extremely important, because it determines its function. In the case of a drug, the structure of the protein that makes up the API determines the therapeutic effect). To date, the structure of 190,000 proteins has been studied. DeepMind (a Google subsidiary company) announced in July 2022 that its artificial intelligence system AlphaFold, predicted the 3D structure of more than 200.000.000 proteins.
Source: https://www.deepmind.com/blog/alphafold-reveals-the-structure-of-the-protein-universe
When it comes to the use of artificial intelligence in the GMP production of medicines, there is still a long way to go. There are multiple possible applications. Many of these have already been suggested, but implementation – in most cases – has not yet been achieved or has been achieved in a few plants only. Some examples are:
- Quality control
- Process optimisation
- Process control
- Predictive maintenance
- Production planning
- Quality control
One of the quality controls performed in the pharmaceutical sector is the condition of the primary packaging. In the case of vials, for example, it would be unacceptable to deliver a medicine to a patient with a broken vial. To avoid such situations, the vials are visually checked by a person or a machine. By means of machine learning, a model can be trained so that this check is optimised to avoid false positives or false negatives during the quality control of the primary packaging.
- Process optimisation
A model is able to analyse the behaviour of a reaction or a process depending on its conditions. This tool can be applied to biopharmaceutical processes to optimise them for higher yields. The model will be able to learn how the process behaves and suggest the best conditions or actions at each stage to improve its performance.
- Process control
In line with the previous point, a process could be controlled autonomously by a model. The model would be connected to a control system capable of making decisions. The data taken from the system would be analysed by the model in real time. The model would process and analyse this data, and then apply a series of actions through the control systems to ensure the maximum performance of the process.
- Predicitive maintenance
Equipment maintenance is a key activity to maximise equipment lifetime and to avoid unexpected production downtime. The frequency of maintenance activities is often not based on data, but on the experience of plant operators, which can lead to errors. Artificial intelligence makes it possible – through data analysis – to predict when an unexpected equipment event is most likely to occur. And thus recommend a maintenance shutdown to avoid such an event. This principle could be applied for example to the frequency of maintenance of a washing machine based on the vibration of the machine.
- Production planning
Production planning can be a real headache, especially for those drugs where demand fluctuates significantly over time. In these cases, a model that analyses past demand data for the same or similar drugs could be used to give an estimate of the batches required. This data could be linked to other data, such as the time of year, temperatures or economic indicators, to provide even more accurate predictions.
The possibilities that artificial intelligence opens up are immense, so much so that we cannot even imagine where it will take us in a few years. We are undoubtedly facing one of the great technological revolutions. And sooner or later it will also reach the biopharmaceutical sector. At Klinea we are eager to see where there is only noise and where we can make real changes with this technology. That’s why we are already preparing and training ourselves to continue helping our customers.
If you are interested in learning more about artificial intelligence and how we can help you, contact us: klinea@klinea.eu
