Thank you for Subscribing to Life Science Review Weekly Brief
Artificial intelligence has great potential to impact the synbio field by creating comprehensive models that reduce the number of experiments needed.
FREMONT, CA:Although artificial intelligence has greater potential to influence the synbio field, it has had a limited impact on synbio until recently. There are successful applications of AI, but they are still limited to particular datasets and research questions. Currently, data mining, statistics, and mechanistic modeling are the primary drivers of computational biology and bioinformatics in the field, and the line between them and AI or machine learning (ML) is often unclear. For example, clustering is a data mining method that identifies patterns and structures in gene expression data, and these patterns indicate whether the engineered modifications lead to a fatal outcome for the cell.
These clustering techniques furthermore serve as unsupervised learning models, finding structure in unlabeled datasets. These methods and novel AI or ML approaches in development will have a much-expanded role and impact in the future of synbio as larger datasets become customarily available.
Transcriptomics data volumes double every seven months, and high-throughput workflows for proteomics and metabolomics are becoming increasingly available. Additionally, the gradual automation and miniaturisation through microfluidics chips of laboratory work hint at a future where data processing and analysis are the primary multipliers in synbio. AI and synbio intersect in various ways, such as applying existing AI and ML to existing datasets, generating new datasets, and creating new AI and ML techniques to apply to new or existing data.
A fundamental challenge in synbio helping AI surmount involves predicting the impact of bioengineering approaches on the host and the environment. Without the ability to predict the bioengineering outcome, synbio’s goal of engineering cells to a specification can only be achieved through difficult trial and error. AI provides an opportunity to use publicly available experimental data to predict the impacts on the host and environment.
Genetic Construct Design for Programming Cells:Many synbio efforts focus on engineering genetic constructs or circuits, presenting very different challenges from designing electronic circuits. The genetic constructs elicit a specific reaction from the cell, much like electronic circuits provide control of an electronic system. DNA can also be synthesised and transferred into cells. This transfer’s global impact on the cellular machinery of the dynamic living organism is neither entirely known nor predictable. In contrast, electrical engineers have the tools to statically design electronic circuit boards to perform a variety of functions and not detrimentally impact the board.
Utilising AI techniques, combining biophysical, machine learning, and reinforcement learning models, to effectively predict the constructs’ impact on the host and vice versa. However, there is a wider opportunity for improvement. For example, different AI techniques are applicable to machine-assisted gene circuit design. This includes expert and multi-agent systems, constraint-based reasoning, heuristic search, optimisation, and machine learning. Sequence-based models and graph convolutional networks have also gained momentum in the engineering biological systems domain.
Factor-graph neural networks integrate biological knowledge into deep learning models. On the other hand, graph convolutional networks predict protein functions from protein-protein interaction networks. Sequence-based convolutional and recurrent neural network models are used to identify potential binding protein sites, genes’ expression, and novel biological constructs’ design. The most useful AI applications will be in comprehensive model development, reducing the number of experiments or designs that must be performed or tested.