Machines such as computers, cars, smartphones, coffee makers, and digital clocks follow algorithms (coding instructions) created by computer science professionals to carry out their specific tasks. Artificial intelligence (AI) algorithms have the goal to make these machines able to mimic the problem-solving and decision-making capabilities of the human mind. In this way, they could not only execute actions smartly but also learn, predict and improve from them.
In medicine, AI algorithms have been increasingly present in softwares to analyze electronic medical records, healthcare management and clinical decision support systems, clinical data such as common tests and medical imaging, genetics and molecular biology research, biomaterials development, among others. In addition, they can be also physically found in medical devices and sophisticated robots used in the delivery of care or in surgery assistance .
Classic AI (also known as symbolic AI) algorithms are based on rules that were hard-coded into models (illustrated below). In this case, humans need to first learn the rules and then code the relationship into an algorithm. Time and resources are then needed to identify and continuously revise and update these relationships. One example of classic AI algorithms is found in clinical decision support systems which contain a knowledge base composed of rules and associations of compiled data. Among other things, these systems can be used to determine drug interaction with the patient. For that, a rule might be that “IF drug A AND drug B are taken, THEN alert the user of contraindications X and side effects Y.” The knowledge base can be continuously updated by users to include information on new drugs .
Machine learning (ML) is a branch of AI that uses statistical learning techniques that allow a machine to establish its own relationship between independent and dependent (response) variables (illustrated above). Algorithms are trained with a dataset to recognize their patterns and create rules that describe their behavior. ML algorithms are able to adapt or improve as training data changes or increases. And they can do that at speeds way higher than human intellect, which has the potential to transform many contemporary research and processes .
During the month of June, we had two great examples of ML algorithms being used as a tool to help research professionals in the formulation of their biomaterials (schematic idea illustrated below). In our recently published application note, we described an internal project in which AI was applied to predict the gelation kinetics of PEGDA hydrogels for a given formulation (PEGDA concentration and light intensity). The ElastoSensTM Bio was used to generate the gelation kinetics database that in turn was used to train and validate an AI model. In the last Article of the month, we have seen another recent study on the use of machine learning to predict the gelation time of silk hydrogels. The prediction helped the authors to more efficiently find formulations that gels within a target window.
As discussed in both publications, the discovery of optimal formulations and the development of biomaterials is hampered by the traditional experimental iterative approach. This is actually a common limiting factor in the biomedical field which oftenly deals with multiple variables and complex responses in excessively large amounts of data. AI has the potential to change the paradigm from data-based experimental science to data-driven predictive science. This would positively impact medicine at many levels from basic research until the arrival of outstanding products and treatments to the patient.
 Hamet, P., & Tremblay, J. (2017). Artificial intelligence in medicine. Metabolism, 69, S36-S40.
 Elbadawi, M., McCoubrey, L. E., Gavins, F. K., Ong, J. J., Goyanes, A., Gaisford, S., & Basit, A. W. (2021). Harnessing artificial intelligence for the next generation of 3D printed medicines. Advanced Drug Delivery Reviews, 175, 113805.
The development of tissue engineering products faces challenges due to the variety of biomaterials available and their dynamic properties once implanted. A recent study used artificial intelligence (AI) to help address this problem. The researchers created a system that automates the prediction of gelation time (when a liquid solution becomes a hydrogel) for different formulations. The hydrogels were composed of silk, horseradish peroxidase (HRP), hydrogen peroxide, and bacterial cells. They utilized differential dynamic microscopy (DDM) to identify gelation time and machine learning (ML) to predict the gelation time of different hydrogel compositions. This allowed them to streamline the process of finding hydrogel formulations with specific properties.
Hydrogels are biomaterials that are widely studied in the biomedical field. They are used, for example, to produce contact lenses and wound dressings, for drug release systems, or as scaffolds for tissue engineering. The design of such hydrogels is often multidimensional since multiple parameters related to their chemical composition and physical properties affect how they are going to behave in vivo.