AI helps scientists discover more biomaterials. This is at least according to an MIT paper released last month that suggests researchers that use AI are surpassing their peers when it comes to research and product innovation.
Researchers sampled scientists working at a large US firm that produces materials for health, optics, and industry. They discovered 44% more materials, filed 39% more patents, and prototyped 17% more new products if they used AI compared to their non-AI assisted colleagues.
Recent advances in deep learning and natural language models have prompted a surge in interest in AI’s potential in product development. The hope is that it will make R&D in material science, including in biomaterials, more efficient.
However, AI and machine learning have been helping discover new biomaterials for years. Here is how AI has already become a key part of generating new sustainable and renewable materials and what new advances are taking shape.
The promise of AI in biomaterials
AI is a computational system that mimics human cognition using machine learning, natural language processing, and other elements. One of its many applications is the prediction, optimisation and design of new materials, including those with molecular structures not found in nature.
AI can give an edge to biomaterials R&D because it takes less time for a computer to generate candidates for materials with a particular set of properties than a human research team on their own.
The difficulty for human researchers is the huge number of possible materials that fit a particular set of criteria. A human must select candidates from this large number of possible materials through trial and error. This roughshod approach risks skipping over promising options or wasting time developing false positives.
Materials are complex, meaning there is sometimes no correlation between certain structures and how they behave on a macro scale that is immediately obvious without using some computational aids for evaluating these connections. AI and machine learning, far more than human labour alone, can more quickly identify promising molecular types and accurately predict a material’s properties based on its structure.
Medical biomaterials
Health and medicine is one of the highest value uses of biomaterials. The earliest biomaterial beneficiaries of AI and machine learning were those used in medicine. These are areas where capital intensive AI approaches to material design can often pay dividends for biomedical engineering companies.
Medical biomaterials have benefited from AI because they often need to have a strict set of criteria to be safe to use in human bodies. One of these is biocompatibility, which means that a body is not likely to reject it as a foreign material.
Biocompatibility is just one property that a medical implant has to meet. How well proteins adhere to their surface is another consideration for these high-end materials – a property known as protein adsorption. QSAR models, an example of machine learning, have been used to predict both protein adsorption and biocompatibility of biodegradable polymers used to reconstruct living tissue.
3D-printed implants, an emerging medical technology, is also set to benefit from AI assisted material design. Material used to 3D print implants is not just biocompatible and protein-attracting, but ‘printable’ too: the ability of a material to be viscous enough to be extruded and dispensed, even when the design of the mould is intricate.
UCL Researchers around Hongyu Chen found that machine learning algorithms were able to successfully predict the printability of different biomaterials, promising to speed up the assessment of materials for bio-printing.
AI in polymers and industrial biomaterials
Medical applications are high value, small-volume uses for biomaterials. However, the field of sustainable design could also profit from AI platforms that can generate new materials hitting the sustainability trifecta: circularity, biodegradability, and made from renewable feedstocks.
AI can be particularly helpful in identifying new biomaterials that meet a strict set of criteria. Mimicking the performance of legacy materials is vital in convincing industry to switch from tried-and-tested, cheaper petrochemicals options for renewable alternatives.
Biobased polymers are a prime area that could benefit from AI-assisted design. These are a wide class of biomaterials with many applications, including as replacements for oil-based plastics.. Due to their complexity and numbers, researchers cannot check through all candidate materials one by one, accurately predicting their properties and behaviours.
In addition to designing the material and predicting its strength, thermal retention, and degradability much more quickly than trial-and-error laboratory tests, AI models can also perform economic assessments that help researchers identify those that would be most efficient and cheapest to produce at scale.
The company TNS offers a custom platform for polymer design, named polySCOUT that is supposed to aid the discovery of sustainable renewable materials. The platform has already helped Senbis develop a biodegradable polyester for textile fibres. The project aimed at mitigating the problem of microplastics and saw Dutch research institutes collaborate with the company on the problem.
Earlier in 2024, an AI polymer information company Matmerize and South Korean biopolymer producer CJ Biomaterials began a material design collaboration. CJ Biomaterials, which recently created a PHA that degrades naturally in soil and oceans, successfully tested Matmerize’s material design AI platform to optimise their newly designed biobased polymers.
Matmerize’s capabilities have been picked up elsewhere with Asahi Kasei Corporation using its informatics platform PolymRize to accelerate R&D in sustainable polymers, including biodegradable textiles.
Generative AI in industrial fermentation
Chat GPT may be the public face of generative AI but the computational structures underlying it – large language models – could also help biomanufacturing by exploring new kinds of genetic edits that could solve particular problems.
Many biobased chemicals and materials are industrially fermented, which means they are formed in the host bodies of functional microbes, harvested, and purified. These industrial microbes are genetically customised so that they produce particular molecules in their body consistently and at scale.
One of the barriers to scaling industrial fermentation is raising the productivity of these host microbes without running into biological tradeoffs (for example, microbes that produce large amounts of target chemicals at speed are also slower to mature, which raises costs).
Researchers could now use what commentators have called ‘ChatGPT for CRISPR’ to find new, counter-intuitive gene editing solutions. Machine learning has been around in industrial fermentation for years but recent leaps in AI – which is capable of generating genuinely new research avenues, not just data analysis – could offer even quicker optimisation.
With models trained on biological data (protein and genome sequences) rather than human language like in proper Chat GPT, new generative AI models could help researchers create super-functional microbes with just the right properties for scaling industrial fermentation.
However, some scientists are concerned that these large and complex new generative models’ workings are so opaque to the human observer that scientists do not gain any proper understanding of the biological mechanisms behind the results. This is why at Oak Ridge National Laboratory, researchers are opting for scaled down, ‘explainable’ AI models that can help guide selection for editing the genome of E. coli, a bacteria used widely in biomanufacturing.
AI confronts climate risks
AI could also speed up the discovery of biobased replacements for synthetic chemicals much more quickly than otherwise. For example, synthetic biologists are starting to develop plant root microbiomes – or communities of microbes – that could support agriculture with fewer petroleum and fossil-dependent inputs.
By using computational techniques combined with genome engineering, researchers are hoping to develop clusters of specific microbial species that can perform specific functions.
These living, custom-built communities could help agriculture become more resilient to the climate challenges ahead. The hope is that they could one-day turn them into responsive units that act on the plant roots in different ways according to changing environmental conditions in the soil, such as drought, high salinity, or heat.
AI helps microbiologists in this field by processing and analysing large amounts of raw data, helping to sort and identify various microbiome communities with key properties that could serve particular agricultural applications. Some recent examples of AI platforms being used to advance the field of smart microbiomes are the open-source DADA2, MicrobiomeAnalyst, and QIIME 2.
Outside the laboratory, AI-aided functional microbiome design is hitting the market through companies like Evogene, a synthetic biology company that offers MicroBoost AI, a discovery and development platform for microbial products. The offerings range from soil microbiomes for use in agriculture to industrial applications.
AI and its potential uses have captured the public imagination since the release of Chat GPT. Yet just as rapidly, serious concerns have emerged about the sustainability of this brave new phase in AI development. The massive data centre capacity needed to support its large sets of algorithms has cast doubt on whether the tool brings net positives for the environment, as some have argued.
However, AI’s use in discovering scaleable, biodegradable, and high-performance materials from renewable feedstock could be one way of repurposing the technology for more sustainable ends. Instead of being directed at water and energy-intensive applications in finance and content creation as it is today, targeted uses in sustainable materials and food security R&D would allow the technology to realise its promise as a tool for bringing tangible social and environmental benefits.