Aptamers, often referred to as ""synthetic antibodies,"" are nucleic acid-based molecules that selectively bind to target analytes in complex biological samples, such as whole blood. They can undergo reversible structural changes upon binding, allowing for real-time detection. By conjugating electroactive reporters to aptamers, these structural changes can be monitored electrochemically. Due to their reagentless nature, these biosensors are highly suitable for both in vitro and in vivo applications. Our lab specializes in aptamer-based sensors and has published several studies on their development and applications. However, their structure-based sensing mechanisms make aptamer responses highly sensitive to environmental factors, including temperature, solution pH, and ionic composition. These variations impact accuracy when mapping signals to target analyte concentrations, even with the use of calibration curves.
On the other hand, computational sensing extracts meaningful information from noisy, complex sensor data while addressing environmental variations. Integrating machine learning and deep learning enhances accuracy and optimizes signal interpretation. Importantly, the ease of arraying aptamer sensors by simply modifying DNA sequences provides a scalable sensing approach. Furthermore, Compute-in-Memory (CIM) boosts energy efficiency by performing multiply-accumulate (MAC) operations for AI inference directly within memory, reducing data movement and power consumption. This enables edge devices to operate more efficiently in applications such as biomedical diagnostics and IoT monitoring.
We propose an aptamer-based computational biosensing technique with CIM to enhance both sensing accuracy and energy efficiency. The system integrates arrays of aptamers along with pH and temperature sensors. We leverage a deep learning model to correct noise from pH and temperature fluctuations, ensuring accurate aptamer sensing while improving energy efficiency.
Project is currently funded by: Federal