Artificial Intelligence (AI), a rapidly advancing field of computer science, focuses on developing software and systems capable of making intelligent decisions through learning and analysis. Over recent years, AI has emerged as a game-changer in material sciences, transforming labor-intensive traditional methods through the introduction of intelligent simulations and automated experimentation[1].
Finding the right probe for the target analyte
For over two decades, in silico computational methods have been instrumental in selecting aptamer candidates tailored to specific targets[2]. These methods utilize structural data to perform simulations and identify aptamers with the highest affinity[3].
With the rapid advancements in AI-driven technologies, researchers and industry players now leverage these tools to predict protein secondary structures, docking outcomes, and aptamer selection processes more efficiently[4].
Using AI can also be ideal for searching large databases of aptamers and identifying aptamer structures that target specific biomarkers. By doing so, it is possible to significantly reduce the time required to identify aptamer structures for selection. In fact, key features such as sequence, structure, and binding affinity serve as critical inputs for AI models.
Material Design and Optimization
AI can also be used for the design and optimization of materials for graphene-based biosensors. Recent breakthroughs in AI applications for graphene[5] include the ability to predict its electrical, mechanical, thermal, and cytotoxic properties, as well as recognizing atomic structures and microscopic geometries. This advancement extends to inverse design, allowing researchers to tailor graphene compositions and microstructures.
Furthermore, AI enhances graphene research by enabling chemical detection, motion recognition, and 3D imaging of graphene composites. In parallel, graphene itself is increasingly utilized in AI hardware, particularly in neuromorphic computing chips, which are seen as potential solutions to overcome Moore’s Law limitations.
AI-based methods are already integral to graphene research, expediting computational processes and supporting diverse applications, as illustrated in Figure 1. Databases like the Computational 2D Materials Database are gaining significant attention for their role in this advancement.
A notable example includes optimizing the fabrication of graphene-based biosensors by targeting specific regions of graphene that influence its affinity with probes, such as aptamers.
Figure 1. Artificial intelligence-driven advancements in graphene applications2
Sensor Calibration and Signal Processing
Transitioning from material sciences to advanced signal analysis, AI techniques can play a crucial role in handling and interpreting complex signals generated by highly sensitive and specific biosensors.
Point-of-Care-Testing (PoCT) is essential in modern healthcare, with machine learning-enhanced biosensors like electrochemical, lab-on-a-chip, and wearable devices improving accuracy, sensitivity, and speed[6] (Figure 2). For example, a notable application integrates AI with DNA-functionalized graphene gas sensors, enabling the detection and discrimination of chemical vapor mixtures in conditions, such as humid human breath. Using feature selection analysis and CNN models (classification), this system demonstrates the synergy of sensor arrays and AI techniques for real-time gas discrimination and early disease diagnosis, as outlined in Figure 3[7].
Figure 2. Overview of AI-ML Assisted PoCT Devices: Integration of wearable, lab-on-chip, electrochemical, and colorimetric sensors with AI/ML to enhance diagnostics, treatment, monitoring, and environmental safety6
Figure 3. AI-integrated DNA-functionalized graphene sensor array for mixed gas detection7
By ‘learning’ from diagnostic labeled data, more accurate “cut-off” thresholds can be determined for the biosensors. This enhances their diagnostic performance, fine-tunes their calibration, and helps distinguish between signal noise and relevant biological interactions, ensuring accurate diagnostics.
Additionally, by using advanced computation, such as AI-driven analytics, the generated values can be better understood, which further supports healthcare professionals’ decision-making in providing care.
Use of AI in 2D-BioPAD
CeADAR brings 11 years of AI services and development experience, specializing in developing AI strategies, prototypes, and solutions that help businesses improve their products or services staying competitive in this fast-changing global marketplace.
In 2D-BioPAD, CeADAR applies its advanced AI solutions to identify aptamer sequences for specific biomarker targets. This helps speed up the aptamer selection process and narrow down the pool of aptamer candidates and ensure that the aptamers integrate well with the graphene-based substrate.
CeADAR carefully chose the best computational methods for aptamer binding research, after comparing various advanced techniques. Using this approach, CeADAR generated a set of candidate aptamer sequences for Aβ and pTau-217 biomarkers, achieving strong interaction scores. The next steps include using AI to optimise the aptamers’ length optimisation and improve the functionalised graphene to enhance the performance of the biosensors, especially in their interaction with magnetic nanoparticles/aptamers in the electrochemical and GFET biosensors.
Additionally, CeADAR is exploring AI solutions to support graphene-based biosensors’ fabrication process, starting with its functionalization based on available online data and input from the technology partners. Other features, such as sensitivity (the lowest detectable concentration), defect levels, and conductivity, can also be optimised.
From aptamer selection to biosensor fabrication, AI is reshaping the future of graphene-based biosensors, driving efficiency and innovation. CeADAR’s expertise ensures the integration of advanced AI strategies for groundbreaking applications in healthcare diagnostics.
Wanna learn more about CeADAR? Click here!
Stay tuned for more.
[1] Pyzer-Knapp, E. O., et al., (2022). Accelerating materials discovery using artificial intelligence, high performance computing and robotics. npj Computational Materials, 8(1), 84.
[2] Rost, B., & Sander, C. (1993). Prediction of protein secondary structure at better than 70% accuracy. Journal of molecular biology, 232(2), 584-599.
[3] Chushak, Y., & Stone, M. O. (2009). In silico selection of RNA aptamers. Nucleic acids research, 37(12), e87-e87.
[4] C Lee, S. J., et al., (2023). Design and prediction of aptamers assisted by in silico methods. Biomedicines, 11(2), 356.
[5] Huang, M., Li, Z., & Zhu, H. (2022). Recent advances of graphene and related materials in artificial intelligence. Advanced Intel. Sys., 4(10), 2200077.
[6]Bhaiyya, M., Panigrahi, D., Rewatkar, P., & Haick, H. (2024). Role of Machine Learning Assisted Biosensors in Point-of-Care-Testing For Clinical Decisions. ACS sensors, 9(9), 4495-4519.
[7]Hwang, Y.J., Yu, H., Lee, G. et al., (2023). Multiplexed DNA-functionalized graphene sensor with artificial intelligence-based discrimination performance for analyzing chemical vapor compositions. Microsystems & Nanoengineering, 9, 28.
